DEPARTMENT OF THE INTERIOR GEOLOGICAL SURVEY
MINUTES OF THE NATIONAL EARTHQUAKE PREDICTION EVALUATION COUNCIL NOVEMBER 16-17, 1984 Menlo Park, California
by Clement F. Shearer1
Open File Report 85-201
This report is preliminary and has not been edited or reviewed for conformity with U.S. Geological Survey publication standards and stratiqraphic npmenclature.
llJ.S. Geological Survey, 106 National Center Reston, Virginia 22092 TABLE OF CONTENTS
page Preface i 1 Minutes of the November 1984 meeting ^ 1. Agenda for November 1984 meeting 12. List of members, National Earthquake Prediction Evaluation Council 13. Appendices: A. The Parkfield, California, Prediction Experiment - W. H. Bakun and A. G. Lindh 16. B. Holocene activity of the San Andreas fault at Wallace Creek, California - Kerry E. Si eh and Richard H. Jahns 56. C. Terms for Expressing Earthquake Potential, Prediction, and Probability- Robert E. Wallace, James F. Davis, and Karen C. McNally 71. D. USGS, Terminoloqy for Geologic Hazards Warnings 79. 11.
PREFACE
The National Earthquake Prediction Evaluation Council (NEPEC) was established in 1979 pursuant to the Earthquake Hazards Reduction Act of 1977 to advise the Director of the U.S. Geological Survey (USGS) in issuing any formal predictions or other information pertinent to the potential for the occurrence of a significant earthquake. It is the Director of the USGS who is responsible for the decision whether and when to issue such a prediction or information. NEPEC, also referred to in this document as the Council, according to its charter is comprised of a Chairman, Vice Chairman, and from 8 to 12 other members appointed by the Director of the USGS. The Chairman shall not be a USGS employee, and at least one-half of the membership shall be other than USGS employees. The USGS has not published the minutes of earlier meetings of NEPEC. This open-file report is the first in an anticipated series of routinely published proceedings of the Council. -1-
NATIONAL EARTHQUAKE PREDICITION EVALUATION COUNCIL Friday, 16. November, 1984 Menlo Park, California
Council Members Present Dr. Lynn R. Sykes, Chairman, Lamont-Doherty Geological Observatory Columbia University Dr. John R. Filson, Vice-Chairman, Chief, Office of Earthquakes, Volcanoes, and Engineering, USGS Dr. Clem Shearer, Executive Secretary, USGS Dr. John Davies, Alaska Department of Natural Resources Dr. Thomas McEvilly, University of Cal ifornia, Berkeley Dr. Mark Zoback, Stanford University Dr. Keiiti Aki, University of Southern California Dr. James H. Dieterich, USGS Dr. William Ellsworth, USGS Dr. Wayne Thatcher, USGS Dr. Robert E. Wallace, USGS Observers Present Paul Seigel, USGS Kerry Sieh, USGS Al Lindh, USGS . Robert Page, USGS George Gryc, USGS Teresa Rodriguez, USGS John Healy, USGS Sandra Schulz, USGS William J. Kockelman, USGS Wanda Seiders, USGS Cynthia Ramseyer, USGS (recording secretary) Edna King, USGS
Opening Remarks Chairman Sykes opened the meeting by reviewing several goals he would like to see the NEPEC pursue: to meet several times a year for on-going reviews of high-risk areas; to advise the Director of the USGS of earthquake hazards and earthquake potential at specific sites; to take a more active role in earthquake prediction; and to publish quickly information (scientific papers and data bases) reviewed by NEPEC regarding earthquake potential and predicition. Filson observed that this meeting could be an historic moment for the NEPEC if it should chose to take a more active role. He suggested that NEPEC can "do more than respond to other people's prediction" and that the NEPEC could set a pace that would enable it to present to the country a "systematic review of the situation in seismic areas." Filson also noted that there is a polarization of opinion about long-term and short-term prediction, and that an issue before NEPEC is how to determine criteria for prediction and for advising the USGS Director of earthquake potential and hazard. Sykes reviewed briefly the history of the NEPEC, which was chartered under the 1978 National Earthquake Hazards Reduction Program, and outlined some of the -2- responsibilities and issues NEPEC might pursue: o need for a mechanism for quick publication of data, o relationship with the media and press, o relationship (reciprocity) with the California Earthquake Prediction Evaluation Council, o short-term responses to rapidly changing conditions, o review of critical seismic areas, and o determination of language for issuing earthquake hazard warnings. Concerns not on this meeting's agenda, but which nevertheless should be addressed: o A review of the legal status of the NEPEC, especially the personal and professional liability of individuals, in both the Federal and private sectors, who serve on the NEPEC; Council may ask for Dept. of the Interior solicitor's review and advice. o How to minimize or avoid misrepresentations to the public. o Ongoing communication within NEPEC by means of small working groups that might meet or teleconfer on a regular basis.
o Guidelines for NEPEC repsonse to predictions that are presented to NEPEC. o Recommendation that NEPEC continue to follow its previous policy of not making any reviews of foreign predictions unless specifically asked by the USGS Director.
AGENDA ITEM: NEPEC's ROLE Wallace agreed that NEPEC should be more active. Also, he stated that earthquake predictions will not be made by individuals but by groups and will evolve with the accumulation and analysis of data. He further noted that the ongoing monthly prediction meetings at the USGS reflect such an evolutionary process, in that it is a group effort between the USGS and private contractors that relies on a redundancy of data. Wallace also noted that raw data and preliminary interpretations of the monthly USGS meetings are on record and printed quickly. Dieterich spoke to NEPEC taking a more active role in long-term forecasting, and the need to identify pre-agreed upon "triggers" that would enable NEPEC to respond quickly in making short-term predictions. He recommended regular reviews of long-term forecasting by the Council as a way to focus the long-and short-term purposes of the NEPEC. -3- Davies spoke to a change of emphasis for NEPEC towards long-term prediction. By so doing, NEPEC could lend credence to the idea that certain circumstances promote further investigation of earthquake potential. Aki, in referring to the Japanese model of earthquake predicition, voiced some concern about the time demands and logistics of calling the NEPEC together on a more frequent or regular basis. McEvilly pointed out that earthquake probability maps and the Parkfield experiment have been developed since the last NEPEC meetings (1/81 and 6/82). FUson pointed out that California earthquake probability maps have not been brought before the NEPEC and have not been formally transferred to the State of California; and that the USGS has previously issued several formal statements regarding earthquake hazards to the State of California and that these statements should be reviewed and either updated or cancelled. Sykes referred to the Japanese model for earthquake prediction and felt that there was a problem with the council in Japan being too large and too formal to make it directly applicable to the United States. The ongoing monthly review at Menlo Park (USGS) could in part suffice as an informal, ongoing mode of earthquake prediction communication and information. Sykes/Filson suggested that specialists be asked to be present at NEPEC meetings to present their data or make evaluations. NEPEC should also be receptive to persons coming before the Council with earthquake predictions; it needs to respond quickly to data that is put before it. Dieterich stated that long-term prediction is necessary: NEPEC must be aware of data on long-term predictions, so as to make an informed judgement about short-term predictions. Ellsworth recommended a NEPEC base of long-term potential data, and a systematic and ongoing review of long-term prediction data by NEPEC, so that if NEPEC is forced to (or decides to) make short-term predictions, the Council has information available on potential long-term behavior. Sykes mentioned that three groups (USGS, Lament, and Caltech) have produced earthquake probability maps that are essentially in agreement with one another: NEPEC should review these yearly or every other year. Zoback mentioned short-term problems confronting NEPEC: 1. NEPEC hasn't yet acquired the background (data) to make judgements about specific sites, and 2. NEPEC needs to familiarize itself with instruments and data related to specific sites. Suggested that the NEPEC needs a system to gather, synthesize, and evaluate data. Sykes responded with a suggestion that to make sure data gets shared quickly and accurately, perhaps data could be kept simultaneously in several locations and be available to qualified users by a dial-up phone capability. -4- Wallace brought up the "Ridgecrest Affair" wherein politics entered into a scientific/technical problem -- this is an issue when the use and application of scientific data comes under political purview and pressure. Thatcher suggested that formal decisions need to be made in a logical sequence: identify areas that need concentrated attention and bring regular updates to NEPEC. Filson felt it was necessary to identify and limit areas of study to specific sites; otherwise NEPEC will be forced into taking a merely responsive rather than active role in earthquake prediction. McEvilly suggested that perhaps NEPEC should remain strictly responsive to prediction data brought to the council, rather than risk a conflict of interest either by reviewing predictions that the Council will have guided by its review process or by soliciting or sponsoring probability studies of certain sites. Filson pointed out that the USGS already has "in house" prediction activities: unusual data should be brought to the Director's attention with NEPEC's evaluation and advice. Aki suggested a prosecuting/defending mode for NEPEC, with the Council acting as a jury, determining the proofs and validity of earthquake predictions. Copies of a new California law (Chapter 1284, Laws of 1984) concerning public immunity from liability for issuing earthquake warnings were distributed. Zoback countered with the observation that cranks are easy to deal with: data are hard to evaluate, endorse, criticize, and/or recommend. Wallace made the following recommendation regarding earthquake predictions: 1. Identify regions of long-term earthquake potential. 2. Limit the areas and/or sites that will be evaluated and reviewed. 3. Get background and information data for high-priorty areas. 4. Be prepared to add new data for a site or area and make a statement. 5. Have pre-prepared statement (similar to SCEPP) regarding earthquake areas, sites, and recommendations. Davies pointed out that by taking an activist role in certain earthquake areas/sites, the NEPEC could compromise its ability to make an impartial judgement. McEvilly agreed and thought that NEPEC should identify itself as either an "impartial jury" or a "activist participant" in the selection of sites and the gathering of data. Sykes pointed out that NEPEC would increase the risk of making poor recommendations if it were not fully informed of all or most of the data. -5- Dieterich wondered what does NEPEC need to do its job? NEPEC needs data to do its job; without it the Council cannot make an intelligent evaluation and decision regarding earthquake prediction statements. Filson: NEPEC should get more actively involved in critical areas, for the purpose of becoming more knowledgeable about high-priority areas and thereby facilitating its role of advising the Director of the LJSGS about "issuing predictions". There should also be a familiarization of the data in areas of long-term concern by the NEPEC members; then if there are new data, the NEPEC could meet to evaluate the new or changed data and then make a recommendation to the Director. The Chairman called for a CONSENSUS of the discussion regarding Long-term Prediction: o The NEPEC will take a more activist role within the confines of the Charter. o The Charter does not need revision at this time. o NEPEC will look at long-range forecasting on a regular basis and will review updated earthquake probability maps. o NEPEC will limit areas/sites that it will examine. o NEPEC will consider predictions brought to it for other areas. o Speakers will be invited to present prepared data on specific areas to the NEPEC for informational purposes; they are asked to leave a 2-page summary and copies of their figures. They are asked, as far as possible, to include estimates of magnitude and probability of any events predicted. o Regarding the press/media, NEPEC meetings can't be kept secret; information regarding the meetings must be given to the press. However, sensitive data can be discussed in closed executive session. If there are implications regarding public safety, the press should be included. Dieterich supported this proposal with the observation that SCEPf (Southern California Earthquake Preparedness Project) meetings are routinely closed, as the press can be intrusive and demanding and can "interfere with the process of reaching a rational decision in a real crisis." Executive sessions could be called. Edna King, (USGS Public Affairs Officer) was consulted, and she observed that cameras can always be excluded, and that there is no need to hold meetings on a "press welcome" basis. It was also noted that at the June 1982 NEPEC meeting, the Council voted to keep its meetings open to the press, with the option to go into executive session if the need arose. Sykes recommended that the "press question" be tabled until the next NEPEC meeting, and that at that meeting, legal representation be invited to address both the "open to the press" question and legal liability of NEPEC members. -6-
Agenda Item: DESIGNATION OF TERMS AND SITES FOR EARTHQUAKE PREDICTION Wallace was asked to present briefly the findings of a paper (co-authored with J. Davis and K. McNally; copies distributed) on earthquake prediction terminology designed to set standards on terms so planning agencies could respond to predictions issued by USGS. The terms below were accepted by the California State Seismic Safety Commission in October of 1983. SUGGESTED TERMS DEFINITIONS Long-term Earthquake Potential No specific time window; earthquake could occur in a framework of decades, centuries, or millenia. Earthquake Prediction A specific time window shorter than a few decades. Long-term Prediction Time window of a few years to a few decades. Intermediate-term Prediction Time window of a few weeks to a few years. Short-term Prediction Time window up to a few weeks. "Watch" and "forecast" are terms that would be used in conjunction with a prediction regarding a specific event, at a specific magnitude, at a specific location and within a specific time frame (i.e. Parkfield, CA). Wallace made a motion that these prediction terms be adoped by the NEPEC. The following discussion ensued: Filson noted that the USGS terms of "notice", "watch", and "warning" have been revised (distributed Federal Register, January 31, 1984) to a one-tier system that encompasses a HAZARD WARNING, defined as a "greater risk than normal and a threat that warrants near-term public response." The USGS makes a probabilistic statement regarding the likelihood of an event, and state and local governments are given notice of hazardous conditions. Wallace noted that various planning communities have done a good job of making specific contingency plans (preparedness plans) for short-term predictions, and that the proposed prediction terms are targeted for positive psychological impact and public response. Filson suggested that extant warning statements (distributed for southern California and Yakataga region in Alaska) issued by the USGS be reviewed and recommendations regarding their present status be made to the Director. Ellsworth raised the question about criteria, or the matrix of probabilities and time windows: high probability/short-term correlation? short-term, low probability of great risk, correlation? Wallace cited NEPEC agreement of warnings being issued on a high probability/short-term/great risk basis. Filson added the suggestion that these terms be recommended to the Director, -7- USGS for use as official USGS terminology. Accepted by consensus.
Agenda Item: POTENTIAL SITES FOR NEPEC REVIEW The following sites/locations were cited by Sykes for NEPEC consideration: 1. Parkfield, CA, and areas located near northern end of 1857 rupture zone 2. Southern California: San Andreas fault from Tejon Pass to Sal ton Sea, Anza gap, northern end of San Jacinto fault 3. Alaska: Shumagin and Yakataga Gaps 4. San Francisco Bay Area, CA: San Andreas fault from near San Jose to SE end 1906 rupture zone, Calaveras and Hayward faults, East Bay area 5. Eastern Sierras: area to the north of 1872 rupture zone; but not including volcanic hazards 6. Wasatch Front, Utah: Salt Lake City 7. Puget Sound, Washington: Pacific NW subduction zone 8. Puerto Rico Trench/Virgin Islands 9. Kern County/Coalinga, CA The following persons addressed the NEPEC on the following sites: Robert Page of USGS discussed the Yakataga Gap in Alaska. John Davies discussed the Shumagin Gap, and made the observation that this site is likely to have a major earthquake in the next two decades, and that it is of concern due to increasing population, economic impact, and off-shore drill ing. Ellsworth discussed California, with the observation that the San Andreas fault in the Salton trough is creeping, and that a large earthquake would not be unexpected in Southern California at any time. High probability areas are Anza, Riverside, and Parkfield. In the San Francisco Bay Area, he mentioned the following as potential sites for study and evaluation: San Andreas (especially San Juan Bautista area), Calaveras, Rogers Creek, San Gregorio, and Hayward faults. Sykes suggested the following two-tiered priority schedule for NEPEC's review over the next one to two years: 1 II Parkfield, CA Eastern Sierras Southern California Wasatch Front, Utah S.F. Bay Area Washington State Alaska Puerto Rico - Virgin Islands Southern San Joaquin Valley, CA After discussion, the following priorities were agreed upon. The next meeting will be devoted to the southern San Andreas fault and to the San Jacinto fault. 1 II Parkfield, Ca Washington State Southern California Wasatch Front, Utah S.F. Bay Area Eastern Sierras Alaska Puerto Rico - Virgin Islands
Agenda Item: Parkfield Evaluation Lindh distributed the paper The Parkfield, California, Prediction Experiment, by W.H. Bakun and A.G. Lindh, and presented slides indicating that Parkfield earthquakes have a mean recurrence time of 21.7 years; probability estimates are very high for an event to occur within the next 10 to 20 years; and an event can be expected to occur approx. 1988 + 4 years (2 standard deviations). Thatcher mentioned the unique opportunity to anticipate and monitor a "predicted" earthquake. Paul Siegel and Kerry Sieh presented information regarding their work and data evaluations regarding the Parkfield site. (A portion of Sieh's paper was distributed to those present). Sieh argued that the northern end of the 1857 rupture zone to the south of Parkfield could break in conjunction with the next Parkfield shock, producing an earthquake of magnitude 7 to 7%. Discussion regarding the Parkfield approach: Aki: good approach, an earthquake will occur there; short-time precursors are being observed and confirmed in the laboratories. Helps fill the gap between long-term and short-term prediction. Present lack of intermediate-term precursors. Wallace: Suggested that the Council set up scenarios to determine how the Council might respond. Ellsworth: Suggested development of a series of discussion points as an opportunity to "capture the earthquake;" e.g., what probability does NEPEC desire for a prediciton? 10%-, 30%, 50% ? This now becomes a question of nerve and will, and commitments to certain thresholds and instrument maintenance to capture and record every possible bit of data. Sykes: Research needed on intermediate-term precursors; establish an observation program for high-risk sites. There is a lack of baseline instruments that are field deployable and limitations on access to private lands for conducting experiments and deploying monitoring instruments and equipment. Dieterich: Noted considerable public education benefits; if USGS/NEPEC are willing to take the risk of making a qualified prediction at Parkfield, which -9- is an optimal location for making an earthquake prediction. Wallace: Recommends that NEPEC make a decision today as a dry-run of a formal warning that Parkfield will have an earthquake. Go public and deal directly with Parkfield's local government. Accompany the warning with probability maps, calculated intensity maps, public education and information, and details on what will happen to structures, and how to prepare for and mitigate risks. He also noted an operational aspect of sustaining the prediction effort and keeping everyone involved on his or her scientific toes. Also, any statement must be accompanied by telling people what to do to mitigate risks, therefore the State Office of Emergency Services must be included in the warning process. Wallace suggested that a press release be issued as soon as possible about a high probability of long-term potential for a Parkfield earthquake. Follow up with ground motion maps, etc. Sieh: suggested that USGS/NEPEC put the public welfare into the picture, and endorsed Wallace's comments.
CONSENSUS was reached on the following: 1. The Parkfield Earthquake Prediction Experiment is one of the highest priorities for the U.S. Earthquake Hazards Program and has the highest probability for a successful prediction. 2. Endorsed the general aspects of the prediction made by Bakun and Lindh that a M 6 earthquake will happen in the Parkfield area by 1988 _+ 4 years. 3. Advise USGS Director to make an information statement regarding the Parkfield earthquake as a long-term prediction. Include the Bakun and Lindh paper in the recommendation. 4. Add to the Parkfield prediction that there is a significant potential of a larger earthquake (M 7.0 to 7.5) growing out of (in conjuntion with) a seismic event in Parkfield, and which may break to the southeast for as much as 25 miles. 5. Work should be undertaken (especially geologic trenching) to explore the prehistoric record at Parkfield. 6. The Open-file report of this meeting should include the Parkfield paper by Bakun and Lindh. 7. Have a short follow-up on Parkfield, including documentation and maps at next meeting. 8. There is a clear need for more attention to intermediate-term precursors. 9. The experiment needs more real-time digital analysis of the seismic data. Water-level data need to be transmitted and analyzed quickly. 10. USGS must overcome internal turf problems in an effort to be more cooperative in sharing and analyzing data (i.e., wells and water levels, -10- putting more earthquake monitoring instruments into existing wells).
Agenda Item: OPTION PAPER FOR EARTHQUAKE PREDICITON STRATEGY by USGS Filson provided background on a draft paper, (copies distributed) explaining that it developed from a request of the Secretary of the Interior for a USGS proposal (including implementation plan and budget for FY 86) before June 1985. The earthquake hazard in Southern Califronia has been a high priority in Washington DC for several years, and Congress has been pushing USGS to become more operational in its approach to earthquake prediction. Proposed options: 1. Continue current research activity. 2. Deployment of strain and seismic equipment in clusters at specific sites of high earthquake potential along the San Andreas Fault. 3. Development of a prototypical earthquake prediction system along the San Andreas Fault in Southern California from Santa Barbara to San Diego: every 20 km along the locked section of the southern San Andreas Fault. 4. Comprehensive employment of 50-60 clusters in Southern California and L.A. basin along the San Andreas Fault and ancilliary faults. Each cluster would cost approximately $2 milllion to install and approximately $300,000 annually to maintain and operate. This option paper has been submitted to the National Academy of Science, the California Seismic Safety Commission, and NEPEC for comment. Earthquake prediction is not a mature science and some concern has been expressed that Southern California is not getting the USGS's best effort. The Department of the Interior feels that public safety is important enough to invest in the work to develop a prediction system for earthquakes. Comments Aki: emphasis appears to be on short-term prediction; an excellent approach. Caveat that some site proposals don't appear to be at nucleation points of likely earthquakes. Wallace: Coherence of instrumentation is a factor for locations: need to find places where there is enough seismic activity to get data. Picking good nucleation points for clustering is a good way to get comprehensive data. Ellsworth: Automated early-warning systems that react to strong ground motion and shut down critical facilities (nuclear power plants, banking computers, etc.) should be allowed for in an overall strategy. Early-warning systems probably belong in the technical strategy. McEvilly: There is an assumption that the USGS knows how to do the job; he is not convinced that the USGS is ready or prepared to undertake this sort of program. Dieterich: Social responsibility indicates that we (USGS & NEPEC) must do the best we can with what's available. -11- Thatcher: Option 4 is a "star wars" approach unless there is a continuous development in logical sequence from Option 1 to Option 4. Davies: Option 4 puts too many resources into one program with analysis of cost/benefit. Dieterich: Pointed out that it will be a political decision.among the present options and that the USGS is not in a position to make the final decision among the options. Davies: Most efficient method may be long-range predictions for hazardous areas and to do pre-event mitigation to lessen earthquake risks rather than put money into "science" that may or may not work. Sykes: Present U.S. program for prediction and hazards reduction is seriously underfunded. Public interest is mainly in prediction, not mitigation. Substantial increase in funding is needed; it is crucial that the research program be expanded. Better data on depth of earthquakes are needed; better M 3-4 data are needed. Wallace: Noted the competition with disaster-relief and disaster-planning agencies for funds; he feels the USGS is underfunded in its earthquake prediction program. Lindh: There is a pragmatic problem with the proposed research in that elaborate instrumentation is a direct result of and dependent upon "brilliant" geophysicists to develop and maintain it. Without trained geophysicists to design, maintain, and interpret data from instrumentation, nothing will happen. Shearer: Suggested that in cooperation with the National Science Foundation, NEPEC could encourage grants and training programs to develop support personnel and instrumentation. Wallace: Need for compatible disciplines, i.e. trenching isn't worth much without dating techniques. Also, research and experimentation needs to be done in the sociology and psychology of automated early-warning systems.
NEXT MEETING: MARCH 1985 Agenda: Review of Legal Liability of NEPEC NEPEC relationship with the media/press Southern California
Minutes taken by: Dr. Clement F. Shearer Cynthia C. Ramseyer, OEVE USGS, National Center, MS 106 345 Middlefield Road, MS 922 Reston, VA 22092 Menlo Park, CA 94025 FTS 928-6208 FTS 467-2313 12
Fall 1984 Meeting National Earthquake Prediction Evaluation Council Friday and Saturday, November 16-17, 1984 Menlo Park, California
Friday Opening remarks, Lynn Sykes, Chairman General discussion on scope of Council and strategy for future, review of terms and types of statements and reports that should be issued by the Council, press relationships. (See Wallace, Davis, and McNally paper.) Designation of areas that should be systematically considered. Review of current efforts and earthquake potential in San Francisco Bay area, southern Alaska, and southern California. (Discussion leaders with knowledge of each of these areas will be present.) Discussion of "Option Paper for Earthquake Prediction Strategy." (Filson can provide background and elaborate on options.) Review of Parkfield prediction experiment (Kerry Sieh and USGS staff). Should Council endorse current Parkfield prediction?
Saturday
Field trip to Parkfield. We shall leave early and return late to Menlo Park. Details at meeting. About a 4-hour drive from Menlo Park, 4-5 hours in the field, 4 hours back. 13
NATIONAL EARTHQUAKE PREDICTION EVALUATION COUNCIL November 1984
Dr. Lynn R. Sykes Higgins Professor of Geological Sciences CHAIRMAN Lamont-Doherty Geological Observatory of Columbia University Palisades, New York 10964 Office: 914/359-2900 Home: 914/359-7428
Dr. John R. Filson Chief, Office of Earthquakes, Volcanoes, VICE CHAIRMAN and Engineering U.S. Geological Survey MS 905 National Center Reston, Virginia 22092 Office: 703/860-6471 Home: 703/860-2807
Dr. Clement F. Shearer Hazards Information Coordinator . EXECUTIVE SECRETARY Office of the Director U.S. Geological Survey MS 106 National Center Reston, Virginia 22092 Office: 703/860-6208 Home: 703/620-9422
Dr. Keiiti Aki Department of Geological Sciences University of Southern California Los Angeles, California 90007 Office: 213/743-3510 Home: 213/559-1350
Dr. John N. Davies State Seismologist, Alaska Department of Natural Resources, Division of Geological and Geophysical Surveys, and, Adjunct Associate Professor, Geophysical Institute, University of Alaska 794 University Avenue, Basement Fairbanks, Alaska 99701 Office: 907/474-7190 Home: 907/455-6311
Dr. James F. Davis State Geologist, California Department of Conservation California Division of Mines and Geology 1416 Ninth Street, Room 1341 Sacramento, California 95814 Office: 916/445-1923 Home: 916/487-6125 14
Dr. James H. Dieterich Research Geophysicist Branch of Tectonophysics U.S. Geological Survey 345 Middlefield Road, MS 977 Menlo Park, California 94025 Office: 415/323-8111, ext. 2573 Home: 415/856-2025
Dr. William L. Ellsworth Chief, Branch of Seismology U.S. Geological Survey 345 Middlefield Road, MS 977 Menlo Park, California 94025 Office: 415/323-8111, ext. 2782 Home: 415/322/9452
Dr. Hiroo Kan amori Division of Geological & Planetary Science California Institute of Technology Pasadena, California 91125 Office: 818/356-6914 Home: 818/796-8452
Dr. Thomas V. McEvilly Chairman, Department of Geology and Geophysics University of California, Berkeley Berkeley, California 94720 Office: 415/642-4494 Home: 415/549-0967
Dr. I. Selwyn Sacks Department of Terrestrial Magnetism Carnegie Institution of Washington 5241 Broad Branch Road, N.W. Washington, D.C. 20015 Office: 202-966-0863 Home: 301/657-3271
Dr. Wayne Thateher Chief, Branch of Tectonophysics U.S. Geological Survey 345 Middlefield Road, MS 977 Menlo Park, California 94025 Office: 415/323/8111, ext. 2120 Home: 415/326-4680
Dr. Robert E. Wallace Chief Scientist, Office of Earthquakes, Volcanoes, and Engineering U.S. Geological Survey 345 Middlefield Road, MS 977 Menlo Park, California 94025 Office: 415/323-8111, ext. 2751 Home: 415/851-0249 15
Dr. Robert L. Wesson Research Geophysicist Branch of Seismology U.S. Geological Survey MS 922 National Center Reston, Virginia 22092 Office: 703/860-7481 703/276-7900 Home: 703/476-8815
Dr. Mark D. Zoback Professor of Geophysics Department of Geophysics Stanford University Stanford, California 94305 Office: 415/497-9438 Home: 415/322-9570 16
APPENDIX A
The Parkfield, California, Prediction Experiment W. H. Bakun and A. G. Lindh
in press, Earthquake Prediction Research, Terra Sci. Pub. Co., Tokyo 17
THE PARKFIELO, CALIFORNIA, PREDICTION EXPERIMENT
W. H. BaKun and A. G. Lindh
ABSTRACT
Moderate-size earthquakes occurred on the Parkfield section of the San Andreas fault in central California in 1881, 19U1, 1922, 1934, and in 1966. The earlier Parkfield earthquakes were similar to the 1966 event, leading to the hypothesis of a characteristic Parkfield earthquake with recurring, recognizable source features. A simple recurrence model that explains most of the historic seismicity near Parkfield implies that the next characteristic Parkfield earthquake will occur within a four year time window centered on 1987-1988. A Parkfield Prediction Experiment, designed to monitor the details of the final stages of the earthquake preparation process is underway. Uoservations and reports of anomalous seismicity and INTRODUCTION Analysis of the probability of damaging earthquakes In California suggest that the Parkfield-to-Cholame section of the San Andreas fault 1n central California Is the most likely site of a damaging earthquake in the next several years (see figure 1). Lindh (1963) found a b?5jproDability of a magnitude 6 earthquake at Parkfield in the next 10 years. Available data suggest that a much narrower time window, Iybb-lyb9, for the occurrence of the next Parkfield earthquake can be established. Since this time window is near, and because historic Parkfield earthquakes have been so similar, Pancfield provides a unique opportunity to prepare in detail an experiment to observe the final stages of the earthquake preparation process. The results of this experiment should provide the understanding of that process so critical to the design of earthquake prediction efforts in other areas. The last damaging Parkfield earthquake, on June 28, 1966, had a Richter local magnitude ML Of 5.6 (Bakun and McEvilly, 1979, 1984) and a seismic moment M of 1.4xl02^ dyne- cm (Tsai and Aki, 1969). Although large enough to cause significant damage if located in a metropolitan area, the shock caused only minor damage to the large cattle ranches and sturdy wood frame homes in the sparsely-populated Parkfiela region. Maximum modified Hercalli intensities of VIII were observed over an area of a few hundred square kilometers centered on Harkfiela and the Cholame Valley. The source of the 1966 earthquake is adequately described for our purposes here by a simple model: unilateral rupture propagation to the southeast over a 20 to 25-km-long section of the San Andreas fault, herein called the rupture 19 locus, between two geometric discontinuities in the fault trace (lindh and Boore, 1981). The northwest discontinuity, adjacent to the epicenter of the 1966 shock, is a 5° change in the strike of the fault trace. The term preparation locus will be used to descrioe the 1 to 2-km-long section of fault that includes both the fault bend and the main shock epicenter. Available data support the view that earlier damaging Parkfield earthquakes were similar to the 1966 event, leading to the hypothesis that Parkfield main shocks have recurring, recognizable source features (Bakun and McEvilly, 1964). Paricfield shocks with these attributes are called characteristic Parkfield earthquakes. Our working hypothesis is that the next damaging Parkfield earthquake will be characteristic, i.e., resembling in detail earlier shocks, In particular the 1966 event for which much detailed information is availaole (e.g., hcEvilly et al., 1967; Brown et al^, 1967), HISTORIC SEISHICITY Parkfield earthquake sequences with moderate-size main shocks occurred on February 2 in 1881, rtarch 21 in 1901, March 10 in 1922, June 7 in 19J4, and June 28 in 1966. Although the Parkfield-to-Cholame section of the San Andreas fault has been tentatively identified as the locus of the probdble epicenter of the 1857 Fort Tejon great earthquake and its two moderate-size foreshocks (Sieh, 1978a), data are not sufficient to constrain slip on the San Andreas fault near Parkfield in 1857 (Sieh, 1978b). Epicenters of one, or both, of the 1857 foreshocks as well as the epicenter of the main shock in 18b7 might lie on the San Andreas fault southeast of the Parkfield-to-Cholame section. 20 The times of Parkfield earthquake sequences, including 1857, are plotted in figure 2 against the earthquake sequence counter; i.e., 1857 is number 1, 1881 is number 2, etc. The time between sequences is remarkably similar, with the mean intersequence time * 21.9^3.1 years. Although the time of the 1934 sequence is not consistent with the regular intersequence interval, the time of the 1966 sequence reestablishes the intersequence spacing in that (1966-1922)72 = 22 years. The two straight lines represent linear regressions of the dates on the counter 1. Using all six dates, origin time * 20.b*I+1837.6 (solid line in figure 2) suggesting that the next Harkfield sequence, i.e. number seven in the series, was due in the spring of 1983. Ignoring the apparently anotnolous 1934 date, origin time = 21.7*I+1b36.^ (dashed line in figure 2), suggesting that the next sequence will occur at the beginning of 1988. Clearly, occurrence of another Parkfield sequence in the next several years would not be unexpected. THE CHARACTERISTIC PARKFIEL0 EARTHqUAKE The 1934 and 1966 Parkfield sequences were remarkably similar. In addition to the common epicenter, magnitude, fault-plane solution and unilateral southeast rupture of the main shocks, identical M. » 5.1 foresnocks preceded each main shock by 17 minutes (Bakun and nctvilly, 197*, 1984). The lateral extent of aftershock epicenters over the rupture locus in 1966 (HcEvilly et ajk, 1967) repeated that in 1934 (Wilson, 19.50). Much less data are available for Parkfield sequences prior to 1934. Nevertheless, most of the datd are consistent with the hypothesis that the 21 earlier main shocks In 1881, 1901, and 1922 were similar to those 1n 1934 and 1966. The eplcentral location of the main shock in 1922 is constrained by the Love-Pn arrival times at Berkeley, CA U * 24ukm) to the 18-km-long section of the fault northwest of the preparation locus {Bakun and ptcEvilly, 1954). The data permit a common epicenter for the 1922, 1934 and 1966 main shocks near the southeast end of the preparation locus. A comparison of seismoyrams for the 1922, 1934 and 1966 main shocks recorded at the same sites (e.g., see figure 3) suggests that within experimental errors (" 10-20%, tne seismic moment MQ in 1922 and in 1934 were each equal to the MQ for 1966 (Bakun and McEvilly, 1984). Although the features of the main shocks are similar, there are notable differences in the foreshock activity (see figure 4). The 1934 main shock was preceded by a nearly 3-day-long foreshock sequence. The 1934 foreshocks included an ML 5.0 foreshock 55 hours before the main shock. Whereas the immediate (17 minutes) ML 5.1 foreshocks in 1934 and 1966 were identical, there was no early foreshock activity in I9bb comparable to that in 1*3^ (see figure 4). There are no reports of felt foresnocks preceding the main shocks in 1881, 1901, or 1922, so that ML & foreshocks probably did not preceed these early events. Furthermore, there are no foreshocks in 19^2 evident on the Berkeley Bosch-Omori seismograms; ML 4 1/2 Parkfield shocks probably would be noticeable on these records. The similarities in the main shocks suggest that the Parkfield-to-Cholame section of the San Andreas fault is characterized by recurring earthquakes with predictable features. The notion of a characteristic earthquake with predictable features means that the design of a prediction experiment can be 22 tailored to the specific features of the recurring characteristic earthquake. Also, as shown in the next section, the hypothesis permits the construction of a recurrence model thalTcan explain most of the historic seismicity at Parkfield. * A Recurrence Model for Parkfield Earthquakes The limited data available on the recurrence of large and great earthquakes along plate boundaries around the world apparently is consistent with a time-predictable model, for which the time interval between successive shocks is proportional to the coseismic displacement of the preceding earthquake (Shimazaid ana Nakata, lybO; Sykes and guittmeyer, 19bl). The fundamental principles of the time-predictable model are contained in Reid's- (1910) analysis of the mechanics of the lyub California earthquake. That is, an earthquake occurs when the strain accumulated since the preceding earthquake results in sufficient stress to rupture the fault surface. Adding the concepts of a constant failure stress threshold, a constant rate of strain accumulation, and variable stress drop results in the time-predictable moael. Unfortunately this simple model is not supported by the data available for the last three Parkfield earthquakes: although comparable coseismic displacements in 1922, 1934, and 1966 are inferred from the observations, the time intervals differ by more than a factor of 2 (12 yrs versus & yrs). However, simple adjustments to the assumptions that drew the time-predictable model from Reid's analysis result in another rnoael that we call the Parkfield Recurrence Model, whic accounts for the historic seismic wA 23 activity at Parkfield. Like the time-predictable model, the Parkfield recurrence model assumes a constant loading rate and an upper bound stress threshold ff( , corresponding to the failure or yield stress of the fault. Whereas the time-predictable model permits variable stress drop, the Parkfield recurrence model assumes a characteristic earthquake (constant stress drop) and permits failure before O"j is reached. Of course such a model is useful in a predictive sense only if these early failures occur infrequently. The Parkfield recurrence model is illustrated in figure 5. The constant stress threshold at which most characteristic earthquakes occur is represented Dy3j". A constant loading rate of 2.8 cm/yr was used to approximate the 3 cra/yr rate of relative plate motion across the creeping .section of the San Andreas fault to the northwest of the Parkfield section (Burford and Harsh, 1980). We assume that the Parkfield earthquakes in Itibl, 1901, 1922, and 1934 and 1966 were identical, with 60 cm of coseismic slip representing a constant average static stress drop of a few tens of bars. A simple physical model can qualitatively account for the features of the Parkfield recurrence model. Let en* the upper stress threshold ^correspond to times when the failure stress is approached generally over the entire fault, at which times failure must occur. That is, there are no late characteristic Parkfield earthquakes. Following Brune (1979), we can devise a triggering scenario that permits the occasional early characteristic earthquake. Consider an asperity, i.e., the preparation locus, adjacent to a weak, creeping fault section, i.e., the rupture locus. If a local stress concentration at the asperity exceeds the failure stress there, then 24 the rupture in a resulting relatively high-stress drop small shock might easily extend into the weak rupture locus and continue until resistance to rupture is sufficient to stop the earthquake (e.y., Husseini e_t £l_., 1975; Das, 1976). (At Parkfield, the geometrical barrier at the southeast end of the rupture locus provides sufficient resistance to rupture to stop the characteristic Parkfield earthquakes.) Thus a smaller Parkfield shock might grow into a characteristic earthquake when the failure stress is approached only locally in the preparation locus. Local, rather than general approach of the failure stress, would correspond to v M, > 4.0 tend to occur at a higher rate afterr duration of the time window, these calculations Imply that the next Parkfield earthquake should occur In 1988.0^ 1.8, i.e., between 1986 and 1989. RECENT SEISMICITY Although earthquakes occur throughout central California, most of the shocks In recent years lie along the San Andreas fault (see figure 6). Not shown here are the sequences of earthquakes east of the San Andreas near New Idria in October 1982 (ML5.4) and near Coalinga in May 1983 (ML6.5). Earthquakes on the San Andreas are shown as a lineation of epicenters 3-5 km southwest of the San Andreas fault trace. This apparent mislocation is presumably the result of lateral variations in crustal velocity not adequately modeled in the location algorithm. Host of the shocks on the San Andreas occur on the creeping section to the northwest of the preparation locus. The section southeast of Cholame that broke during the great Fort Tejon earthquake of 1857 is currently locked, with no measureable fault creep and only infrequent small shocks. A cross section of the seismicity along the fault (figure 7) Illustrates the predominance of the activity to the northwest of the preparation locus, defined by the locations of the main shock and the immediate MLS.I foreshock in 1966. This activity northwest of the preparation,locus is concentrated at focal depths less than about b km. Focal depths of the main shock and the immediate foreshock in 1966 are about 8 km (Lindh H-i]^, !9fc3), deeper than most of the events to the northwest of the preparation locus and deeper than the majority of aftershocks in the rupture locus (see figure 7)). The recent clusters of seisiaicity within the 19o6 aftershock zone (shaded area in figure 7) occur at the concentrations of aftershocks identified by Eaton et £l_. (1970), 27 Prominent features of the seismicity near the 1966 hypocenter are Illustrated in the schematic cross-section shown in figure 8. Since 1975 a number of magnitude 4 to 5 earthquakes have occurred near the preparation locus. This is the seismicity that, according to the Parkfield recurrence model shown in figure 5, occurred at o"greater than the second stress threshhold Vg. The 1934 and 1966 Parkfield sequences were proceded by >\5.1 foreshocks located at the northwest edge of the preparation locus. The immediate foreshocks had larger stress drops than had other ML«, earthquakes that occurred in the area in the past 50 years (tiakun and HcEvilly, 1981). These other ML5 earthquakes all occurred a few kilometers northwest or southeast of the preparation locus (tfakun aria i-icEvilly, 1981). It is not clear whether the larger stress drops of the immediate foreshocks result from their location at the edge of the preparation locus or because they preceded their respective main shocks by only 17 minutes. Note that the early MLS foreshock located 2 kilometers northwest of the preparation locus that preceded the 1934 earthquake by 55 hours was a relatively low stress drop source (Bakun and McEvilly, 1981). A magnitude 4 earthquake in June 1982 near the same location and the magnitude 5 shock in Septei.ioer 1975 located t> km northwest of the preparation locus were lower stress drop sources as well (O'rJeili; 1984; Bakun and McEvilly, 1981). btress drops for a nuiiioer of smaller earthquakes that have occurred near the preparation locus indicate a similar spatial pattern (see figure 9). Lower stress drop sources tenu to occur around the higher stress drop sources. Note that the focal depths of the main shock and immediate foreshock in 1906 are relatively uncertain so that the hypocenters of these events whose epicenters define the extent of the 28 preparation locus might lie within the group of higher stress drop sources shown in figure 9. The implication is that the preparation locus is characterized by relatively high stress drop sources, whether or not the sources are foreshocks. Under this interpretation, the immediate foreshocks in 1934 and in 1966 were relatively high stress drop sources because of their location at the edge of the preparation locus rather than because they immediately preceded the main shocks. The historic seismicity suggests that the preparation locus is critical in the nucleation of characteristic Parkfield earthquakes. The last two characteristic earthquakes, in 1934 and in 1966, were preceded by foreshocks within the preparation locus. These events, like other shocks within the preparation locus, are relatively high stress drop sources, consistent with the notion that the 5° bend in the fault at the preparation locus is the point where stress is concentrated. Clearly any earthquakes located in tne preparation locus, or any other anomalous behavior there, might be precursors to the next characteristic Harkfield earthquake. SEISrtIC INSTRUMENTATION The seismic instrumentation now deployed near Parkfield (see figure lo) is focused to monitor the details of seismic activity in and near the preparation locus. Eleven seismographs of the U.S. Geological Survey's (USGS) central California seismic network (CALNET) are located within a few focal depths of the preparation and rupture loci. In addition, ten Terra-Technology digital event recorders are deployed in a temporary network near the 29 preparation locus; these temporary stations are being replaced by the more-reliable 3-component low-gain CALNET stations. The dense seismograph coverage around the preparation locus should provide documentation of any seismic precursors to the next Parkfield characteristic earthquake. In addition to the seismograph networks, nearly 50 SMA-1 strony-motion accelerographs are deployed near the rupture locus (see figure lu). Tne conception and design of this strong-motion network was a cooperative effort of the USGS and the California Division of rtines and Geology (CDMG). The network is operated ana maintained oy the CiMi. A much sparser strong-motion network was operated near the southeast end of the rupture locus during the 1966 sequence of earthquakes (Murray, 1967) by the U.S. Coast and beodetic Survey and the California Department of Water Resources. Data recorded by that network was the basis of important research on the focal mechanism of earthquakes and the interpretation of near-field strong motion recordings (eg., Aki, 1968; Haskell, 1969; Boore et aj_. 1971; Linah and Boore, lybl). rthile data from that earlier sparse strong-motion network stimulated much discussion, it left unresolved some important questions. In particular, the location of the southeast end of the rupture locus in 1966 is uncertain; the current strong-motion network shown in figure 10 is designed to provide definittve answers to some of these questions. STRAIN il£ASUR£i4ti4TS Reports consistent with signficant precursory aseismic slip along the rupture locus in 1966 provide a strong incentive to deploy strain-medsuriny 30 instrumentation near the rupture and preparation loci. An irrigation pipeline tnat crosses the main trace of the San Andreas in the rupture locus near creepmeter XCK (see figure 11) oroIce and separated aoout 9 hours before the occurrence of the main shock in 1966. Brown et al_. (1967) attribute the break to 1-2 feet of southeast movement of the northeast end relative to tne southwest end. This movement is consistent with the rignt lateral strike-slip displacement across the fault observed in the 1966 afterslip (drown et al., 1967) and on creepmeter recordings near Parkfield since the early 1970s (Burford and Harsh, 1980). However, the time history of the movement tnat resulted in the broken irrigation pipe is unknown; perhaps only a small fraction of the postulated 1-2 feet of displacement occurred in the days and weeks just before the 1966 earthquakes. Also of interest are the reports of very fresh appearing en echelon cracks observed in the rupture locus near creepmeter XUK (see figure 11) twelve days before the 1966 earthquakes (Brown et. al., 1967). (Note that cracks tend to appear each spring in the Cholame Valley IK. Burford, personal communication, 1982) as the clay soil desiccates following the winter rains.) Tne discovery of the cracks in June 196o by delegates to the Second United btates-dapan Conference of Research Related to Earthquake Prediction led to the deployment of a microearthquake study in the area on 16-19 June 19bb, eiynt days oefore the 1966 sequence began; a 24-hour record from that study shows no identifiable magnitude >_ U earthquakes witnin 24 km (Alien and Smitn, lybo). Thus, if of tectonic origin, the en echelon cracks resulted from aseismic slip or fault creep in the rupture locus. The occurrence of l-2cm of fault creep, inferred from the en echelon cracks, would be 4-8 times the annual creep rate 31 at Parkfield. An optimistic interpretation of the broken irrigation pipeline and the fresh en echelon cracks described above is that significant anomalous precursory fault creep occurred at least in the rupture locus in the days and weeks just before the 196b earthquake. If comparable aseismic slip precedes the next Parkfield earthquake, the strain measuring instruments deployed along the rupture locus (see figure 11) will provide clear precursory signals that might be used to issue a short-term prediction. Six creepmeters (see Burford and Harsh, 1980) span the main trace of the San Andreas fault in the rupture locus. Signals from these sensors are recorded on site and also are telemetered to the U.S.G.S. analysis facilities in Menlo Hark, California. Line lengths will be measured each night on a two-color laser distance measuring instrument located at the center of the radial array shown in Figure 10; this instrument provides long term repeatability at the 10"^ level on lines of 3-8 km length. The two-color laser project is a cooperative effort of the University of Colorado and the U.S. Geological Survey. Two Sacks-Evertson volumetric borehole strainmeters are now installed near tne southeast end of the rupture locus (OGH in figure 10); the borehole strainmeters have a sensitivity better than 10"10 and are isolated from first order' surface noise sources such as rain and temperature. The borenole dilatometer project is a cooperative effort of the Carnegie Institute, Washington, D.C., and the U.S. Geological Survey. The two-color laser geodimeter and borehole strainmeter observations should provide corroborative evidence of changes in seismicity and/or creep rate. On a more fundamental basis, they provide the means to define any tectonic deformation leading up to the next characteristic Parkfield earthquake. 32 DISCUSSION Although our understanding of Parkfield earthquakes is far from complete, the available information summarized in this paper suggest some guioelines for short-term prediction of tne next characteristic Parkfield earthquake. SCENARIO 1; FUKESHOCKS IN THE PREPARATION LUCUS, FAULT CKtEK IN Th£ RUPTURE LOCUS. Based on the observations in 1966, we might expect significant foreshock activity in the preparation locus in the hours and minutes before the next characteristic shock and perhaps significant fault creep in the rupture locus in the weeks and days before the event. If such precursors occur, then the current deployment of instrumentation shown in figures 10 and 11 should unambiguously capture the short-term precursory signals and might provide sufficient evidence to support a short-term prediction. SCENARIO 2: NO FORESHOCKS, NO FAULT CREEP IN THE RUPTURE LUCUS. According to the Parkfield recurrence model shown in figure 5, the occurrence times of the Parkfield sequences in Ib81, 19ul, 1922, and I9bt> were not anomalous. While the 1966 event was preceded by significant foreshock activity, the absence of reports of felt foreshocks in lt$bl, 19ol, and 1*22 suggests that these events were not preceded by HL 5 foreshoctcs. Whereas the evidence for significant precursory fault creep in the rupture locus before the 1966 event is ambiguous, there is no information at all concerning analogous changes before the 1881, 1901, or 1922 earthquakes. Clearly tne worst short-term prediction scenario - no foreshocks and no fault creep - would probably lead to the occurrence of the next chdracteristic shock without a short-term prediction. Note however that the epicenter of the main shock in 1922 occurred near 33 the preparation locus. It seems reasonable to assume that some precursory changes, albeit without ML £ 4 1/2 foreshocks, occurred near the preparation locus in 1922. Under the characteristic earthquake hypothesis, the epicenter of the next characteristic Parkfield earthquake will be located near the preparation locus. Hence precursory changes, with or without foreshocks, in the preparation locus are likely. Whereas the two-color laser and dilatometers are favorably sited to detect deformation along the rupture locus, they are relatively insensitive to strain or creep in the preparation locus. Thus, if the only precursors are less-than-gross deformations in the preparation locus (scenario 2), the current instrumentation would likely fail to provide evidence of that deformation sufficient to permit a snort-term prediction. Additional strain-measuring instrumentation near the preparation locus would significantly increase our ability to detect precursors in tne worst-case short-term prediction scenario of no foreshocks and no significant fault creep along the rupture locus. SCENARIO 3: EARLY (1934-LIKE) OCCURRENCE. Scenarios 1 and 2 dealt with circumstances likely to precede a characteristic Parkfield earthquake in 1986-1989, i.e., when & I Z5; . The next characteristic Parkfield earthquake might occur early, i.e. at TF < O] , as in 1934. Could such an ' «7 earthquake be predicted. Unfortunately, data from only one such occurrence, in 1934, is available to address that question. Fortunately, the foreshock swarm in 1934 was so pronounced and prolonged (see figure 4) that it would be easy to recognize a repeat of the sequence of events in 1934, even if no precursory fault creep occurred in the rupture locus. Note the failure of isolated ML b Parkfield shocks in 1939, 1956, and 1975 (see figure 4) to be 34 followed by early characteristic Parkfield earthquakes. This admittedly limited data set suggests that not only are early characteristic Parkfield earthquakes preceded by significant prolonged foreshock activity, but that M^ 5 Parkfield earthquakes either isolated in time, e.g., 1939 and 1956 in figure 4, or only followed within a few hours by small aftershocks, e.g., 197b in figure 4, are not sufficient in themselves to warrant the short-term prediction of a characteristic Parkfield earthquake, uf course the next characteristic Parkfield earthquake can only be early by at most 3 or 4 years in contrast to the 10-year-advance of the 1934 sequence; perhaps tne sequence of events in 1934 cannot be used to anticipate the circumstances preceding a characteristic earthquake early by only a few years. SCENARIO 4: A CHARACTERISTIC PARKFIELD EARTHQUAKE TRIGGERS A LARGER SHUCK. Scenarios 1, 2, and 3 describe circumstances that might precede the next characteristic earthquake, i.e., an M, 5.6 shock bound by the geometrical barriers at the ends of the rupture locus. In this final scenario, we consider the situation where the characteristic earthquake breaks through the right-step en echelon offset at the southeast end of the rupture locus and continues southeast along the San.Andreas fault, growing into a major earthquake. Mechanisms for rupture continuing through an unoroken, or broken, asperity have been developed by Das and Aki (1977). Alternatively, the characteristic earthquake might stop at the echelon offset, ana, in analogy to the triggering mechanism of the early ML *>.o foreshock in 1934, increase the right-lateral shear stress on tne fault southeast of the rupture locus so that another shock eventually starting there would rupture to the southeast. The latter case has been suggested (Sien, 1978a; Lindh ana boore, 35 1981) as the triggering mechanism for the great Fort Tejon earthquake of Ittb7. How might scenario 4 be discriminated in advance: Clearly this scenario presents technical, social, and political problems of the most serious nature. Slip in 1857 along the 50-km-long section of the San Andreas southeast of Cholauie was about 3 1/2 m, significantly less than the y m offset further to the southeast (Sieh, 1978b). Continuation of a Parkfield earthquake to the southeast might result in a rupture length of aoout 9u km and offsets of about 3 1/2 m to the southeast of Cholame (Sieh and Jahns, 1984). Such an event would perhaps be as large as surface-wave magnitude HS 7 1/2 (Sieh and Jahns, 1984). Social and economic consequences of such an earthquake would certainly be more severe than for the characteristic Parkfield earthquake considered in the first three scenarios. Since the average Holocene offset rate across the San Andreas fault at Wallace Creek is 3.5 cm/yr (Sieh and Jahns, 1984), it seems likely that the 3 1/2 m of slip In 1857 largely has been recovered so that the possibility of an earthquake breaking this segment must be taken seriously. Unfortunately, there is little data available to suggest what precursors might discriminate scenario 4 from scenarios 1, 2, or 3. rtodels of rupture through asperities (e.g., Jas and Aki, 1977) suggest that minor differences in the stress field near the asperity, the strength of the asperity, and tne dynamic stress ahead of tne rupture could all be important. Although foreshocks and/or deformation at the southeast end of the Parkfield rupture zone might portend a shock significantly larger than a characteristic Parkfield earthquake, there is certainly no evidence that such need be the case. 36 References Ak1, K., Seismic displacements near a fault, J. Geophys. Res., 73, 5339-5376, 1968. Alien, C. R., and S. W. Smith, Pre-earthquake and post-earthquake surficial displacements, l£ Parkfield earthquakes of June 27-29, 1966, rtonterey and San Luis Obispo Counties, California - preliminary report, Bull. Seism. Soc. Am. 56, 966-967, 1966. Bakun, W. H., and T. V. McEvilly, P-wave spectra for M, foreshocks, aftershocks, and isolated earthquakes near Parkfield, California, bull. Seism. Soc. Am. 71. 423-436, 1981. \ Bakun, W. H., and T. V. rtctvilly, Recurrence:> ~ models1-:-? ->ana Parkfield, California, earthquakes, J. Geophys. Res. 89, Mn PHJGS, 1984. Bakun, W. H., and T. V. HcEvilly, Earthquakes near Parkfield, California: comparing the 1934 and 1966 sequences, Science, 2u5, 1375-1377, 1979. Boore, D. M., K. Aki, and T. Todd, A two-dimensional moving dislocation model for a strike-slip fault, Bull. Seism. Soc. Am, 61_, 177-194, 1971. Brown, R. D., Jr., J. G. Yedder, R. E. Wallace, £. F. Roth, R. F. Yerkes, R. 0. Castle, A. 0. Waananen, R. W. Page, and J. P. Eaton, The . Parkfield-Cholame California, earthquakes of June-August 1966-iurface geologic effects water resources aspects, and preliminary seismic data, U.S. Geol. Survey Prof. Paper 579, 6b pp., 1967. Brune, J. N., Implications of earthquake triggering and rupture propagation for earthquake prediction oased on premonitory phenomena, J. Geophys. Res. 84, 2195-2198, 1979. 37 Buhr, G.S., and A. G. Lindh, Seismicity of the Parkfield, California, region 1969 to 1979, U.S. Geol. Surv. Open-File Report 82-205, 89 pp., 1982 Das, S., A numerical study of propagation and earthquake source mechanism, ScD. Thesis, Massachusetts Institute of Technology, Cambridge, 1976. Uas, S., and K. Aki, Fault plane with barriers: a versatile earthquake mouel, J. Geophys. Res. 82, 5658-5670, 1977. Eaton, J. P., M. E. U'Neill, and J. N. Murdock, Aftershocks of the l*bb Parkfield-Cholame, California, earthquake: a detailed study, Bull. Seism. Soc. Am., 6£, 1161-1197, 197U. Haskell, N. A., Elastic displacement in the near-field of a propagating fault, Bull. Seism. Soc. Am., !59 865-9U8, 1969. Husseini, M. I., D. B. Jovanovich, M. J. Randall, and L. B. Freund, The fracture energy of earthquakes, Geophys. J. 43, 367-385, 1975. Lindh, A. G., M. E. O'Neill, W. H. Bakun, and 0. B. Reneau, Seismicity patterns near Parkfield, California, (abs.): Earthquake Motes 54, 61, iy«3. Lindh, A. G., and D. M. Score, Control of rupture by fault geometry during the 1966 Parkfield earthquake, Bull. Seism. Soc. Am. 71, yb-il6, 1981. Lindh, A. G., Preliminary assessment of long-term probabilities for large earthquakes along selected fault segments of the San Andreas fault , system in California, U.S. Geol. Surv. Open-File Report B3-63, 14 pp, 1983. Murray, G. F., Note on strong motion records from the June 1966 Parkfield, California, earthquake sequence, dull. Seism. Soc. Am 57, 125y-l26b, 1967. rtcEvilly, T. V., W. H. Bakun, and K. B. Casaday, The Parkfield, California, earthquakes of 1966, Bull. Seism. Soc. Am., 57, 1221-1244, 1967. 38 O'weill, H. £., Source dimensions and stress drops of small earthquakes near Parkfleld, California, Bull. Seism. Soc. Am. 74. 27-40, 1984. Reid, H. F., The California earthquake of April 18, 1906, II, mechanics of tne Earthquake, Carnegie Inst. of Washington, Washington, O.C., 1910. Shimazakl, K. and T. Nakata, Time-predictable recurrence model for large earthquakes, Geophys. Res. Lett., 7, 279-282, 1980. Sieh, K. E., Centaral California foreshocks of the great 1857 earthquake, Bull. Seism. Soc. Am., 68, 1731-1749, 1978a. Sieh, K. E., Slip along the San Andreas fault associated with tne great 1857 earthquake, Bull. Seism. Soc. Am. 68, 1421-1448, 1978b. Sieh, K. E., and K. H. Jahns, riolocene activity of the San Anareas fault at Wallace Creek, California, Geol. Soc. Am. Bull., in press, 1984. Sykes, L. R., and R. C. Quittmeyer, Repeat times of yreat earthquakes along simple plate boundaries, Third Maurice Ewing Symposium of Earthquake Prediction, £, edited by U. W. Simpson and P. U. Richards, AbU, Washington, D.C., 1981. Thatcher, W., Seismic triggering and earthquake prediction, Nature, 299, 12-13, 1982. Tsai, Y. B., and K. Aki, Simultaneous determination of the seismic moment and attenuation of seismic surface waves, Bull. Seism. Soc. Am., b9 275-287, 1969. Wilson, J. T., Foreshocks and aftershocks of the Nevada earthquake of December 2U, 1932 and the Parkfield earthquake of dune 7, 19J4, dull. Seism. Soc. Am., 26, 189-194, 1936. 39 Figure Captions Figure 1. Annual earthquake probabilities for selected segments of the San Andreas fault system 1n California (Taken from Lindh, 1983). These estimates are preliminary and should only be used to obtain an overview of the relative earthquake likelihood for different individual fault segments. Figure 2. Series of earthquake sequence at Harkfield since ItiSU (taken from Bakun and McEvilly, 1984). Solid line is the linear regression of the time of the sequence using the last six sequences. Dashed line is the linear regression obtained without the 1934 sequence. The anticipated time of the seventh, i.e., the next, Parkfield sequence for the two regressions is 1983.2 and 1988.U. Figure 3. Surface waves recorded on the Oe Bilt, the Netherlands, east-west (UdN-EW) and north-south (utfN-uS) component balitzin seismographs for the 1922, 1934, and 1966 Parkfield events (taken from Bakun and HcEvilly, Iyti4). Amplituoe and time scales are constant. Brackets indicate the Love- and Rayleigh-wave phases. 40 Figure 4. Parkfield seisraicity relative to the origin times of ML 5 shocks in 1934, 1939, 1956, 1966, and 1975. The times in 1934 are relative to the origin time of the early HL 5.0 foreshock; felt foreshocks in 1934 for which ttuhr and Lindh (1982) assign no magnitude are shown as ML 3 events. Except for the aftershock sequences in 1934 and 1966, no known ML > 3 Parkfield earthquakes occurred within several days of the 75-hour-long time intervals shown. Figure 5. The Parkfield recurrence model. "8\ represents the failure stress of the fault. Constant 2.8cm/yr loading rate and 60cm coseismic slip for the Parkfield earthquake sequences in 1881, 1901, 1922 and 1934 and 1966 are assumed. According to the model, the next Parkfield sequence is expected in 19b8 ^ 2 yr. \ > 4.0 shocks since 1930 are shown at bottom. ML > 4 shocks tend to occur when the stress exceeds^. Figure 6. Earthquake epicenters for 1969-1981 and the location of permanent seismographs in central California relative to geologic features. Most of the area shown is blanketed by Cretaceous and Tertiary marine sediments. Large outcrops of Franciscan melange (Fr) of Mesozoic age are shown, as is the western edge of the San Joaquin Valley, marking the boundary between Tertiary sediments and Quaternary alluvium. Symbols refer to the earthquake focal depths I..., 9, A, tt, ...for..., 41 9-10 km, 1U-11 km, 11-12 km,...). Symbol size is proportional to magnitude (see key). Epicenters were obtained using a one-dimensional crustal velocity model; the band of epicenters located on the San Andreas fault are displaced 3-b km to the southwest because the higher crustal velocity suutnwest of tne fault are not properly accounted for in the location procedure. Priest Valley (PRI) operated by the Univerity of California / Berkeley Seismographic Station and the CALNET station at Gold Hill (GUH) were seismograph stations installed before the 19bb Parkfield sequence. Figure 7. Cross section of the seismicity along the San Andreas fault near Parkfield for the years 1975-1980. The hypocenter of the main shock and the ML$.I immediate foreshock in 1966 are shown as stars. Symbol size is proportional to magnitude. KO vertical exaggeration. Figure 8. Schematic cross section of seismicity (ML > 3) along the San Andreas fault near Parkfield for Iybi>-lyb3. wo vertical . . exaggeration. The shaded vertical band corresponds approximately to the location of a 5! bend in tne surface trace of the fault. The preparation locus is inferred to lie within the shaded region between the hypocenters of the main shock and the MLS.I immediate foreshock in 1966 (the two stars). The aftershocks in 1966, i.e., tne rupture locus, lie southeast of 42 the preparation locus at depths shallower than tt-lu km. Mnce 1975, ML 3.5 earthquakes have occurred near the preparation locus; these sequences are shown toyether with estimates of their source dimensions based on aftershock locations. Figure 9. Cross section along the San Andreas fault zone near Parkfield showing the distribution of static stress drops for a number of earthquakes in 1977-1982 (taken from U'Neill, 1964). The numbers next to the symbols are stress drops in bars. The hypocenter of the main shock and the M, 5.1 immediate foreshock in 196b are shown as filled circles. Focal depths of the 1966 shocks are uncertain to witnin 1-2 km so tnat tneir hypocenters might easily coincide with tne locus of greater stress drop sources shown as filled triangles. Figure 10. Seismograpn and accelerograph deployment along the Parkfield-to-Cholame section of the San Andreas fault relative to the preparation locus and rupture locus of the cnaracteristic Park field earthquake. The epicenter of the 1966 main shock is shown as a star. The location of the southeast end of tne rupture locus is problematic; in 1966, numerous aftershocks and surface cracks were observed over the 2u-km-lony section (cross hatching) immediately southeast of the preparation locus. Surface cracks and some small aftershocks were observed over a 15-km-long section further to the southeast. 43 Figure 11. Strain-measuring Instrument deployment along the Parkfield-to-Cholame section of the San Andreas fault relative to the preparation locus and rupture locus of the characteristic Parkfield earthquake (see caption for figure 10). Names of sites of invar-wire strainmeters, bubble-level tiltmeters, Sacks-Evertsen dilatometers and creepmeters begin with S, T, D, and X respectively. Creepmeter XMrt is located at tne epicenter of the 1966 main shock. TABLE 1. ML >^ 4 Earthquakes Near Parkf ield (1930-1983)* 44 ORIGIN TIME f YEAR MO-DAY HR-MIN(OCT) LATITUDE LONGITUDE ML (ON) (°W) 1 1934 06-05 21-48 35048.0' 12Q020.0' 5.0 1934 06-05 22-52 35048.0' 120020.0' 4.0 2 5.1** 3 1934 06-08 04-30 35048.0' 120020.0' 4 1934 06-08 04-47 35048.0' 120020.0' 5.6*** 5 1934 06-08 05-42 35048.0' 120020.0' 4.5 6 1934 06-08 09-30 35048.0' 120020. O1 4.0 7 1934 06-08 23-23 35048.0' 120020.0' 4.0 8 1934 06-10 08-03 35048.0' 120°20.0' 4.5 9 1934 06-14 14-55 35048.0' 120°20.0' 4.0 10 1934 06-14 15-54 35048.0' 120°20.0' 4.0 11 1934 06-14 19-26 35048.0' 120°20.0' 4.5 12-02 16-07 35058.0' 120°35.0' 4.0 12 1934 4.7** 13 1934 12-24 16-26 35056.0' 120°29.0' 14 1935 01-06 04-04 35056.0' 120°29.0' 4.0 15 1935 10-22 18-37 35055.0' 120°29.0' 4.0 16 1937 02-20 09-58 35056.0' 120029.0' 4.0 17 1938 11-22 15-30 35052.7' 120028.13' 4.2 18 1939 05-02 18-49 35059.2' 120°21.28' 4.0 19 1939 12-28 12-15 35058.17' 120024.62' 5.2 20 1941 12-22 00-54 35056.0' 120029.0' 4.0 21 1942 10-31 10-51 360Q1.86' 120025.71' 4.0 22 1953 05-28 03-51 35057.0' 120028.98' 4.3 23 1953 06-22 15-22 35055.9' 120025.8' 4.4 24 1954 03-09 19-55 360QO.O' 120020.0' 4.0 25 1956 11-16 03-23 35057.9' 120025.7' 5.0 26 1956 12-11 10-56 35056.6' 120028.0' 4.0 27 1958 09-01 11-31 36006.0' 120029.91' 4.6 28 1961 07-31 00-07 35049.4' 120015.8' 4.7 29 1961 12-14 11-51 36000.0' 120030.0' 4.0 30 1966 06-28 04-08 35056.6' 120030.5' 5.1 31 1566 06-28 04-26 35056.0' 120029.6' 5.6 32 1966 06-28 04-28 35055.9' 120029.6' 4.5 33 1966 06-28 04-32 35048.9' 120016.8' 4.0 34 1966 06-28 04-34 35048.9' 120016.8' 4.0 02-19 35055.8' 120027.5' 4.0 35 1966 ' 06-29 4.9** 36 1966 06-29 19-53 35056.8' 120028.6' <*i 35052.0' 120021.5' 4 .2 37 1966 06-30 01-17 f\ 38 1966 10-27 12-06 35056.9' 120041.4' 4 .2 39 1967 07-24 07-08 35055.7' 120026.25' 4.1 35051.2' 120023.09' 4.2 40 1967 08-12 18-57 ** 41 1967 12-21 23-58 35045.3' 120026.8' 4 .3 42 1967 12-31 23-48 35055.31* 120027.15' 4.5 11-17 35056.78' 120030.90 4.4 43 1975 01-06 4.9** 44 1975 09-13 21-20 35059.54' 120033.22' 45 1977 01-24 18-05 35047.23' 120020.96' 4.0 46 1977 11-29 16-42 35056.51' 120029.59' 4.1 47 1977 12-28 02-59 35048.49' 120021.89' 4.0 48 1982 06-25 03-58 35058.32' 120031.38' 4.0 Events for 1930-1979 taken from Buhr and Lindh (1982). Locations for early events are approximate. Data for 1980-1983 taken from preliminary USGS earthquake catalogs. ** ML taken from Bakun and McEvilly (1981). ,v,,« and MrEvillv (1984). 45 CURRENT ANNUAL PROBABILITY (%/o) GORDA BASIN MENDOCINO M7 2%/yr Characteristic earthquakes and cumulative 30 year probabilities ara shown for some named fault segments - MOJAVE M 7.6 - 8 ( 40% ) ] OLEMA M8(3%) U. S.G.S. HAZARD WATCH MAMMOTH - MONO LAKES HAYWARD M 6.5-7 ; S. F. PENINSULA SOUTH Hj2o*j. M7(8%) SAN JUAN BAUTISTA M 6.5 (47%) CREEPING M7?(3%) PARKFIELD M6(99%) CARRIZO MOJAVE M 7.5-8 (40%) INDIO Great M7.5-8 Historic (24%) California Earthquakes 46 2000 r- 1988.0 . 1983.2 1950- 1900 - 1850 345 EARTHQUAKE SERIES r 47 RAYLEIGH LOVE 48 1934 5- * Main Sho 4- m _ Aftershock 3- Sequence I I | II I I 1939 5- 4- 3- 1956 5- M, 4- 3- 1966 5- Main Shock 4- Aftershock Sequence 3- 1975 5- 4- 3- ' II -10 10 20 30 40 50 60 TIME C HOURS P. Slip ( era ) o> * r* _ 000° I I I I -I 857 I 88 I --I90I 1922 I930n 1934 1950- 1966 1970- 1988 ± 2 yrt 990J 6t7 50 36*30' N JOAQUIN DIABLO VALLEY RANGE A (in A ISM - IS70 SIICE 1170 O 3 CONPOIIEIT STkTIOl SE cf M L 0 I966 Aftershock Zone ) d O* (P O 3 O 2 20 o 1 SE MAGNITUDE 1234 . » o AFTERSHOCK 1966 1975 O 1976-1978 1980-1981 D 1982 X 1969-1975 2km 1 53 35*58'______35*56' I T I November 1977 - June 1982 Dcr 28 JUN66,M5 Parkfield foreshock 28 JUN 66,M5.5 Parkfield earthquake 4r- II- 0. 9 A a O D I2i- 21 A 7 G 14 1 I t i 02468 NW SE DISTANCE ALONG FAULT TRACE (km) 54 Preparation Locus a o ParKfleld Rupture Locus Explanation Cholome SMAI * -$- *£* O O Digital Event Recorder -<})-CALNET 3-component O CALNET High-Gain Vertical m 55 N TGH 1 XGH Explanation O Creep Dilatometer Strain 5 km XWT Q A Tilt ' ' Two-color laser 56 APPENDIX B Holocene Activity of the San Andreas Fault at Wallace Creek, California Kerry E. Sieh and Richard H. Jahns reprinted from the Geological Society of America Bulletin, vol.95, no.8 with permission of the author 57 Holocene activity of the San Andreas fault at Wallace Creek, California -KERRY E. SIEH Division of Geological and Planetary Sciences, 170-25 California Institute of'Technology, Pasadena, California 91125 RICHARD H. JAHNS* School of Earth Sciences, Stanford University, Stanford, California 94305 ABSTRACT Finally, we note that the Wallace Creek difficult without greater understanding of its be slip rate is appreciably lower than the average havior during the past several millennia. Wallace Creek is an ephemeral stream in rate of slip (56 mm/yr) between the Pacific In this paper, we present and discuss the geo central California, the present channel of and North American plates determined for logic history of Wallace Creek, a locality about which displays an offset of 128 m along the the interval of the past 3 m.y. The discrep halfway between San Francisco and Los An San Andreas fault. Geological investigations ancy is due principally to slippage along geles that contains much information about the have elucidated the relatively simple evolu faults other than the San Andreas, but a Holocene behavior of the San Andreas fault tion of this channel and related landforms and slightly lower rate of plate motion during the (Fig. la). For the purpose of determining rates deposits. This history requires that the aver Holocene epoch cannot be ruled out. of slip in Holocene time, the channel of Wallace age rate of slip along the San Andreas fault Creek offers excellent possibilities. The channel has been 33.9 ± 2.9 mm/yr for the past 3,700 INTRODUCTION crosses and is offset along a well-defined, linear yr and 35.8 + 5.4/-4.1 mm/yr for the past trace of the San Andreas fault in the Carrizo 13,250 yr. Small gullies near Wallace Creek California has experienced many episodes of Plain of central California (Fig. Ib). It is rela record evidence for the amount of dextral slip tectonic activity during the past 200 m.y. During tively isolated from other large drainages, and, during the past three great earthquakes. Slip the past 15 m.y. horizontal deformations due to therefore, its history is not complicated by in during these great earthquakes ranged from the relative motion of the Pacific and North volvement with remnants of other drainages that -9.5 to 12.3 m. Using these values and the American plates have been dominant. On land, have been brought into juxtaposition. average rate of. slip during the late Holocene, the major actor in this most recent plate-tectonic The simple geometry of Wallace Creek sug we estimate that the period of dormancy drama has been the San Andreas fault, across gests a simple history of development. Arnold preceding each of the past 3 great earth which -300 km of right-lateral dislocation has and Johnson (1909) inferred 120 m of offset on quakes was between 240 and 450 yr. This is accumulated since the middle Miocene (Hill and the San Andreas fault, because the modern in marked contrast to the shorter intervals Dibblee, 1953; Crowell, 1962,1981; Nilsen and channel of the creek runs along the fault for (-150 yr) documented at sites 100 to 300 km Link, 1975). about that distance. Wallace (1968) also in to the southeast. These lengthy intervals sug The San Andreas fault traverses most of ferred a simple history of offset involving inci gest that a major portion of the San Andreas coastal California, running close to the populous sion of a channel into an alluvial plain, offset of fault represented by the Wallace Creek site Los Angeles and San Francisco Bay regions 250 m, then channel filling and new incision will not generate a great earthquake for at (Fig. la). Its historical record of occasional great across the fault. The latest dextral offset of 128 least another 100 yr. The slip rate determined earthquakes (Lawson and others, 1908; Agnew m then accumulated. These interpretations are at Wallace Creek enables us to argue, how and Sieh, 1978) amply demonstrates that it verified and quantified by us in this paper. ever, that rupture of a 90-km-long segment poses a major natural hazard to inhabitants of STRATIGRAPHY AND northwest of Wallace Creek, which sustained these regions. The future behavior of the San GEOMORPHOLOGY as much as 3.5 m of slip in 1857, is likely to Andreas fault thus has long been a topic of great generate a major earthquake by the turn of interest to Californians. Interpretations of histor Figure 2 is a geologic map of the Wallace the century. ical, geodetic, and geologic data have yielded Creek area that is based upon surficial mapping In addition, we note that the long-term estimates of one century to several centuries for and study of sediments encountered in numer rates of slip at Wallace Creek are indistin the time between great earthquakes along the ous excavations. The map shows four main geo guishable from maximum fault-slip rates fault in the San Francisco Bay region (Reid, logic units: older fan alluvium (uncolored), estimated from geodetic data along the creep 1910; Thatcher, 1975). Geologic data indicate younger fan alluvium (green), high-channel al ing segment of the fault farther north. These that similar recurrence intervals apply in south luvium (dark orange), and low-channel allu historical rates of slip along the creeping ern California (Sieh, 1978b, and in press). vium (light orange). A mantle of slope wash and reach thus do represent the long-term that The behavior of the San Andreas fault during local alluvium, which is extensively burrowed is, millennial average, and no appreciable the past few thousands of years is one of the best by rodents, overlies most of the deposits. This elastic strain is accumulating there. clues to its future behavior. Useful forecasts con unit has been mapped (brown) only where it is cerning the likelihood or imminence of a great thicker than -1 m and does not cover units and "Deceased. earthquake along the fault will be much more relationships that need to be shown on the map. Geological Society of America Bulletin, v. 95, p. 883-896, 11 figs., 3 tables. August 1984. 58 SIEH AND JAHNS Figure 1. a. Wallace Creek (WC) is along the San Andreas fault (SAP) between Los Angeles (LA) and San Francisco (SF), in the Carrizo Plain of central California, b. This oblique aerial photograph shows the modern channel, which has been offset -130 m, and an abandoned channel that has been offset -380 m. An older abandoned channel, indi cated by white arrow at left, has been offset f^t^'i^r.-*"'*-- '~ »^^ v,- Xx.' - '>-.--- -475 m. Photograph by R. E. Wallace, 17 September 1974. View is northeastward. FIGURE 2 EXPLANATION Older Fan Alluvium UNITS Underlying all other units exposed at the site, I Hi I Low-channel alluvium there is a late Pleistocene alluvial fan deposit fHhl High-channel alluvium derived from the Temblor Range to the north east. This deposit, here termed the "older fan iHsl Slope wash (mantles mast of area, but mapped only where alluvium," consists of thin sheets, lenses, and boundaries are distinct) stringers of indurate silty clay, pebbly sandy F71 Younger-fan alluvium (dots indicate edges of individual lobes) clay, and sandy gravel. Most of the trenches (Figs. 2 and 3) exposed this unit. Southwest of fPol Older-fan alluvium the fault, the older fan alluvium is covered by various deposits, but northeast of the fault, the SYMBOLS deformed fan surface is incised. Contacts (solid where geomorphically apparent or exposed in Charcoal disseminated within the older fan trench, dotted where buried, dashed where inferred alluvium 4 m below the surface of the fan in Faults (as above; hachures on downthrown side; trench 5 (Fig. 3), yielded an age of 19,340 ± 0.3' 1.000 yr B.P. (Table 1). The lack of major un numbers indicate height of scarp) conformities and paleosols in the older fan - - Selected small gullies offset ~Sm in 1857 alluvium below or above this dated horizon im 5 fbackhae plies that all of the exposed 13 m of the unit Trenches \ ^ ... no ^bulldozer formed during the late Pleistocene epoch. Evi dence discussed below supports a conclusion Crests of small fans and source gullies offset ~9m in 1857 that the fan surface on the northeast side of Landslide, showing headscarp, scarp height, and direction of the fault had become inactive by about 13,000 movement yr B.P. 59 500 figure 2. Geologic map of Wallace Creek. Contours of topographic base map show elevation (in feet) above sea level. 60 Trench 4 sw 19,340 ±1000 14 C yrs. B.R Columnar Section Yrs B.P. Low-chonnel alluvium Trench 2 NE 5m 3,000- 4,200- High-channel alluvium 5,800- 13.250H650. yrs. B.R 10,000- Slope wash (colluvium) _ Younger fan alluvium 13,000 Figure 3. Records of selected excavations at Wallace Creek, California. See Figure 2 for locations. Detailed logs of these and other excavations are available upon request from the author. Older fan alluvium 19,000- HOLOCENE ACTIVITY. SAN ANDREAS FAULT 61 TABI 1 I R-U5KX -\RBON ANAl VKf-S was incised by the creek prior to A.D. 1908 (Sieh. 1977, p. 61). Sample n. > M( < alkxh t nu Wash a'-c . uirrtxted C'alendru. aut Ur Bl' i age; (U BP i GEOLOGIC HISTORY tt(-^ I W 5fT 18.750 450 2237 18.782 450 19.340 100(1 UC-2 i tt-s:: 12.865 I65H 13.250 165" The evolution of Wallace Creek has been 225* \s C -f I \\-SM 5.040 IhO 2151 5.096 5.845 290' rather simple. It is divisible into four periods, U( -" l'\V.5(is 3.730 170 2240 3.772 4.190 270' u c-9-: UW-744 3.720 140 105 3.956 4.480 each of which ends in a sudden change of chan UC-' I V\ -563 3.495 110 -1971 3.580 WO.' UV -563b 3.285 100 - 1971 3.370 nel configuration. McKS-l UW-355 3.340 - 120 155" WC- 1 0-1 UW-742 3.32C - 100 -154 3.476 3.780 \VC-M-I UW-745 I.I 10 180 10.8 1.341 1.035 235* Accumulation of Older Fan Alluvium and Initial Entrenchment of a Channel ,nii' all date uncertainties in the table are at 95'. confidence level At denned b\ Slutver and Polach (1977. p. 3561 As explained in Stuiter and Polach 119771. (.(intentional ages muM be corrected Tor isoiopic fractmnalion Corrected usine half-life of 5.730 \r: uncertainly estimated follow ing suggestion or Klein and others (1982. p 117) Prior to initial incision of Wallace Creek, dur 'From Klein and others (1982. Table 21 ing the late Pleistocene epoch, the older fan allu *Fr«mStimer. 1982 vium gradually accumulated as broad, thin beds on an alluvial fan or apron that extended southwestward from the Temblor Range across Younger Fan Alluvium these sands and gravels residing in a 3- to 4-m- the San Andreas fault (Fig. 7a). The lack of deep channel cut into colluvium. Like their cor small channels within the older fan alluvium in Southwest of the fault (Fig. 2), there is a lo- relatives northeast of the fault, these beds exhibit dicates either that any scarps that formed along bate deposit that we have termed the "younger major episodes of scour and fill. A radiocarbon the fault during this interval were buried before fan alluvium." This deposit overlies and is less analysis of organic matter from trench 10 they accumulated even 1 m of height, or that indurated than the older fan alluvium. It is a yielded an age of 3,780 ± 155 yr B.P. This sam they faced mountainward and served to pond well-sorted gravelly sand with a distinctive im ple was collected from a colluvial wedge in the the older fan alluvium on the upstream side of brication of pebbles that indicates southwest- middle of the deposits in the abandoned channel, the fault. About 13,000 yr B.P.. the first major ward current flow. The unit is thickest near and its age indicates that the abandoned-channel entrenchment of the older fan alluvium occurred trenches 2, 9. and 10 and thins to the northwest, deposits are contemporaneous with the high- (Fig. 7b). Several small gullies were eroded into southeast, and southwest. The boundary of this channel deposits across the fault and upstream. the fault scarp, and their debris, the younger fan composite alluvial fan is inferred from the Figure 5 includes a profile of the high terrace. alluvium shown in Figure 2, was deposited at topography and the trench exposures. A radio The height of the high terrace above the modern the foot of the scarp. At about the same time, the carbon date from charcoal in the upper centi channel is greatest at the fault; the terrace merges initial entrenchment of Wallace Creek occurred. metre of the older fan alluvium (trench 2) with a low terrace ~1 km upstream from the The downstream segment of this initial channel indicates that the younger fan alluvium began to fault. Judging from the elevation difference of now lies outside the mapped area, -475 m accumulate 13,250 ± 1,650 yr B.P. (Table 1). the high terrace across the fault, vertical slip dur northwest of Wallace Creek (beneath the white ing the past 3,800 yr is 3 m, which is a mere arrow at left margin of Fig. Ib). High-Channel Alluvium 2.3% of the horizontal slip during that time period. It is worth noting that within 1 km to the Initial Offset of Wallace Creek Nestled within the channel of Wallace Creek northwest and to the southeast, this vertical slip and Re-entrenchment above the modern stream bed, there are numer diminishes to zero and reverses sense. ous remnants of an ancient terrace (Fig. 2). This After -100 m of right-lateral slip had been surface is referred to as the "high terrace," and it Modern-Channel Alluvium registered by the features formed -13,000 yr is underlain by sand and gravel beds character B.P., the initial downstream segment of Wallace ized by scour-and-fill structures, which we refer Younger sand and gravel beds very similar to Creek was abandoned, and a new segment was to as the "high-channel alluvium" (trench 5 in the high-channel alluvium have been deposited cut, so that a straight-channel configuration was Fig. 3). The massive and poorly sorted nature of in the modern channel of Wallace Creek (Fig. 2, restored across the fault (Fig. 7c). This new some of these "high-channel" beds indicates that and trench 5 in Fig. 3). Like the high-channel segment is the one labeled "abandoned channel" they are debris-flow deposits. Other beds that alluvium, this "modern-channel alluvium" also in Figure 1. « are well sorted and laminated must have been exhibits scour-and-fill structures and interfmgers transported as bedload in the waters of Wallace with debris derived from the channel walls. More Offset and Re-entrenchment Creek. In trench 5, the base of the modern-channel of the Channel Radiocarbon analyses (3) of charcoal from alluvium is 2.5 m beneath the creek bed, and within the high-channel deposits in trench 5 along the entire channel, there is a low terrace For several millennia the newly re-entrenched demonstrate that these beds were accumulating that occurs 1.5 m above the modern creek bed Wallace Creek served as a narrow conduit for through a period from 5845 ± 225 yr B.P. to (Fig. 5). This terrace represents the highest level materials being transported fluvially out of the 3680 ± 155 yr B.P. (samples WC-3, WC-6, and reached by the modern-channel deposits; it nearby Temblor Range. The depth of initial in WC-7 in Table I). formed and was incised within the past 1,000 yr, cision of this channel is poorly constrained, but Southwest of the San Andreas fault, the high- as indicated by the radiocarbon date of 1035 ± it cannot have been more than 1 2 m, which is channel deposits occur in the abandoned chan 235 yr B.P. on charcoal 2.5 m below the terrace the depth of the base of the high-channel depos nel of Wallace Creek (Figs. 1 and 2). Trenches surface in trench 11 (Fig. 6). An early photo its below the surface of the old alluvium in 2, 7, and 8 (Fig. 3) and 9 and 10 (Fig. 4) show graph of the channel shows that the low terrace trench 5. As slip accumulated along the San SIEH AND JAHNS 62 Southeast wall of trench 9 YOUNGER FAN ALLUVIUM Southeast wall of trench 10 Figure 4. Trenches 9 and 10 reveal the various deposits of the abandoned channel. Bt and 62 are scarp-derived breccias. Ct, 2, and 3 are fluvial sands and gravels. Solid triangles indicate location of charcoal that yielded date for channel deposits. An'dreas fault early during the Holocene epoch, was eroded away. About 3800 yr B.P., after the the maximum depth of the new channel is only Wallace Creek developed a bend along the fault channel had been offset -240 m, critical 8.5 m below the top of the high terrace in that reflected the offset accumulated since en changes began to occur within the channel. For trench 5. trenchment (Fig. 7d). Water and debris flowing reasons that we do not understand, debris began within the channel were diverted to the right at to accumulate in the channel to greater thick Offset and Future Re-entrenchment the fault, flowed along the fault for a distance nesses than ever before (Fig. 7e). Trench 5 re of the New Channel equal to the accumulated offset, and then were veals that at the right bend, the accumulation « diverted left and away from the fault. These two was at least 5.5 m deep. Trenches 2, 9, and 10 The new channel has been offset -130 m bends in the channel will be referred to hereinaf show that this accumulation all but filled the subsequent to its creation about 3800 yr B.P. ter as the right bend and the left bend. channel at the right bend. This filling set the (Fig. If). The modern-channel deposits have ac Trench 5 indicates that by -6000 yr B.P., 3.0 stage for abandonment of the channel down cumulated in the new channel during this period to 3.3 m of sediment had been deposited within stream from the right bend and re-entrenchment of time. They are now -2.5 m thick at the right the channel at the right bend. Trench 5 also of Wallace Creek straight across the fault bend and more than 2.5 m thick at the left bend. shows that, locally, at least 1.5 and perhaps 3.3 (Fig. 7e). The new channel was cut no more Although the active channel floor is now m of these high-channel deposits subsequently than 8.5 m below the level of the old channel, as -3 m below the crest of the channel bank at the 63 HOLOCENE ACTIVITY. SAN ANDREAS FAULT -2350 older UJ CO -2300 -SAN ANDREAS FAULT Top of fluvial o CO -2250 LONG PROFILES OF THE WALLACE CREEK DRAINAGE 1500 1000 500 0 (feet) 500 1000 1500 ' I i i 250 0 (meters) 250 DISTANCE FROM FAULT Figure 5. Stream profiles of the modern and the abandoned channels of Wallace Creek. High terrace, indicated by dotted lines, and top of high-channel alluvium, indicated by solid and dashed lines, are offset ~3 m vertically. Figure 6. Trench 11 exposes the upper 2.5 m of low-channel deposits in the modern channel. Solid lines are contacts of individual fluvial beds. Dotted lines represent locally visible layering within these beds. right bend, older modern-channel deposits form discharge. Such a re-entrenchment would estab has accumulated since that entrenchment, one a low terrace surface that is only 1 m below the lish the creek once again straight across the fault. can calculate rather precisely the rate of slip for crest of the bank there. Mr. Ray Cavanaugh, the San Andreas fault. That rate is 33.9 ± 2.9 SLIP RATE OF THE who farms at Wallace Creek, reported to us that mm/yr, and its derivation is explained in detail SAN ANDREAS FAULT water actually spilled over the edge at the right below. bend in the winter of 1971-1972 or 1972-1973. Slip Rate during the Late Holocene The offset of the modern channel of Wallace It is not hard to envision a third entrenchment of Creek is 128 ± 1 m. This figure is obtained by Wallace Creek (Fig. 7g), given another metre or Knowing the date of the most recent en extrapolating the southwestern edge of the two of channel filling and a moderately high trenchment of Wallace Creek and the offset that abandoned channel (labeled 1 in Fig. 8) to its 64 SIEH AND JAHNS intersection with the fault and then measuring the distance from that intersection to the inter section of the modern channel edge (labeled 2 in Fig. 8) with the fault. The same value is ob tained if one measures the distance between the offset segments of the modern channel (labeled 3 b. and 4 in Fig. 8). In making the latter measure 19,300 yrs ago 13,250 yrs ago ment of offset, it is important to realize that the outside edge of the left bend has been eroded by flood waters that have swept against it as they have passed around the left bend. The right bend has not been eroded in this manner, because it is refreshened each time the fault slips. The fact that feature 1 and feature 3 (Fig. 8) intersect the fault at almost the same point strongly suggests that the abandonment of the high channel and entrenchment of the modern channel were contemporaneous. This coinci C. dence also indicates that the new channel was about 10,000 yrs ago 3,700 yrs ago cut straight across the fault without any initial nontectonic deflection of the stream along a fault scarp. The absence of any initial, nontec tonic deflection is also confirmed by the fact that the modern channel is entrenched through a broad topographic high immediately down stream from the fault (consider contours in Fig. 2). If the channel had been deflected along a fault scarp, one would expect it to have cut through a low point on the downstream side of e. the fault rather than a high point. The measured 3,700 yrs ago present separation of 128 m thus is ascribable entirely to tectonic offset. The youngest date from the deposits of the abandoned high channel (3680 ± 155 yr B.P.) provides a maximum age for the modern chan nel, because all of the high-channel sediments were deposited before the modern channel was cut. AH offset of the modern channel thus oc .'. .?££-' ""y^-' '-^\ , ; "^-"'ffi- " -;j curred between this date and A.D. 1857. The average slip rate, therefore, can be no slower than 35.7 ± 1.9 mm/yr [128 ± 1 m/3,680 ± 155-93 yr]. (93 yr is the time between A.D. future 1950, which has been designated zero B.P., and A.D. 1857.) Figure 7. The Holocene-Iate Pleistocene evolution of Wallace Creek. An aggrading "older Additional considerations are necessary to alluvial fan" during the period including 19300 yr ago progressively buried small scarps provide an upper limit to the slip rate. For this formed along the San Andreas fault (SAP) during major strike-slip events (a). Right-lateral constraint, trenches 9 and 10 (Fig. 4) are useful. offsets accumulated during this period, but no geomorphologically recognizable offsets began The high-channel deposits here consist of three to form until 13,250 yr ago, when the "older alluvial fan" became inactivated by initial distinct units, labeled Cl, C2, and C3, that*re- entrenchment of Wallace Creek (b). At this time, erosion of small gullies to the right present three distinct scourings and fillings. The (southeast) of Wallace Creek also resulted in deposition of the "younger fan alluvium" uppermost sediment of channel C2 in trench 10 downstream from the fault. These features then began to record right-lateral offset, and scarps contained the radiocarbon sample the age of began to grow along the fault. About 10,000 yr ago, a new channel was cut across the fault at which is 3780 ± 155 yr B.P. At the time of Wallace Creek, and the initial channel, downstream from the fault, was abandoned (c). The deposition, Cl, C2, and C3 in trenches 9 and 10 new channel remained the active channel of Wallace Creek during the early and middle must have been at or northwest of the right bend Holocene, during which -250 m of slip accumulated (d). This channel filled with "high- of Wallace Creek. Trench 9 is now 145 m channel alluvium" 3,700 yr ago, and Wallace Creek cut a new channel straight across the northwest of the right bend, and so no more fault (e). Between 3,700 yr ago and the present, this youngest channel has registered 128 m of than 145 m of dextral slip has accumulated since right-lateral offset (f). Aggradation of this channel, accompanied by continued offset, will channel C2 was filled 3780 ± 155 yr B.P. probably lead to its abandonment and the creation of a new channel, cut straight across the The trend of C2 between trenches 10 and 9 fault (g). suggests that the edge of C2 actually intersects 65 HOLOCF.NE ACTIVITY. SAN ANDREAS FAULT Put in a different way, the 3,680-yr date may be centimetre or two beneath the base of the fan, any fraction of a recurrence interval younger suggests that a range fire occurred just prior to than the beginning of a strain accumulation deposition of the fan. The charcoal certainly cycle. The beginning of the loading cycle corre would have been oxidized if it had not been sponding to the earliest increment of the 128-m buried deeply very soon after its formation. Ero offset thus may be any time between 3680 yr sion of the fan materials from their source within B.P. and 3680 plus one recurrence interval. As is the burned area may have been a direct result of seen below, the average recurrence interval here the fire, which removed protective vegetative is -310 yr. or 8% of the time between A.D. cover from the ground surface. The charcoal age 1857 and 3680 yr B.P. The actual slip rate thus of 13,250 ± 1,650 yr B.P. thus represents the age Figure 8. The edge of the abandoned chan could be as much as 8% lower than that just of the basal unit of the overlying alluvial fan. nel (1) intersects the fault 128 ± 1 m north calculated, or 32.5 ± 1.5 mm/yr. The late Holo- If the source of the younger fan sediments west of the intersection of the modern cene slip rate thus could be any value between were Wallace Creek, the fan would be offset a channel edge (2) and the fault. The offset of 32.5 ± 1.5 and 35.3 ± 1.5 mm/yr. This range is mere 128 m. This would imply that the fault the modern channel (from point 3 to point 4) conveniently expressed as 33.9 ± 2.9 mm/yr. was inactive between about 13,000 yr and about is also 128 m. These provide the best measure 3700 yr B.P., because we have just shown that of offset during the past 3,700 yr. Slip Rate since 13,250 yr B.P. 128 m of slip has occurred since about 3700 yr B.P. Such a long period of dormancy along the the fault at least 10 m closer to the modern right An additional determination of slip rate along San Andreas fault seems very unlikely to us. and bend. In support of this, we note that the chan the San Andreas fault at Wallace Creek comes so we seek a source for the younger fan that is nel is ~5 m wide and rests entirely southwest of from the 475-m offset of a 13,250-yr-old alluvial farther to the southeast. the fault in both trench 9 and trench 10. Suffi fan from its source gullies. This provides an av The volume of the fan is -25,000 m 3. Candi cient channel width to accommodate a similar erage slip rate of 35.8 + 5.4/-4.1 mm/yr. which dates for the source gully (or gullies) must have deposit southwest of the fault in the modern is not appreciably different from the late Holo- total eroded volumes at least as great as this and channel does not exist until at least 15 m down cene rate of 33.9 ± 2.9 mm/yr. preferably somewhat larger, because some of the stream from the modern right bend. There, the The 13,250-yr-old alluvial fan constitutes the material transported out of the source region crest of the channel bank is ~5 m southwest of "younger fan alluvium" mapped in Figure 2. must have been carried beyond the alluvial fan the fault trace, and, were the channel to fill this The fan radiated from a point that is now lo as suspended load and bedload. year, a 5-m wide deposit analagous to the chan cated very near the modern left bend of Wallace Given this constraint, only two plausible nel fill in trenches 9 and 10 would be deposited. Creek. Its existence is reflected in the bulging of sources for the fan exist within 1 km of Wallace From trench 9 to this geometrically analagous the 2,240-, 2,250-, and 2,260-ft contours toward Creek. The first is a solitary channel -730 m point in the modern channel (labeled 2 in Fig. 8) the southwest (Fig. 2). Even though it is now southeast of the fan apex (E in Fig. 1). This is -130 m. It seems, therefore, that no more buried by 1.5 to 2 m of unmapped slope wash channel originates in the Temblor Range but than 130 m of dextral slip accumulated between and bioturbated materials, the bulging of the drains a much smaller area than Wallace Creek. 3780 ± 155 yr B.P. and A.D. 1857. This yields contours and measurements of thickness in If this is the source, an average slip rate of -63 an upper limit of 35.3 ± 1.5 mm/yr [130 trenches 2, 3,4, and 6 enable construction of the mm/yr for the period 13,250 to 3700 yr B.P. is m/(3,780 ±155-93 yr)] for the slip rate. This isopach map of the younger fan alluvium shown calculated: maximum limiting rate is indistinguishable from in Figure 9. (730- 128)m the minimum limiting rate of 35.7 ±1.9 mm/yr Trench 2 (Fig. 3) exposes the sediments of the = 63 mm/yr. (13,250-3,680) yr determined previously and independently. The fan near its apex. There, the sediments constitute rate must therefore be 35.3 ± 1.5 mm/yr, which a 1.3-m-thick bed of well-sorted, imbricated This would indicate fluctuations in slip rate of at includes the highest maximum value (35.3 + 1.5 sandy gravel. The gravel is composed of tabular least several centimetres per year during the past mm/yr) and the lowest minimum value (35.7 - pebbles of diatomaceous Tertiary marine mud- 13,000 yr, because the average rate for the past 1.9 mm/yr). stone. Imbrication of these tabular pebbles 3,800 yr has been -34 mm/yr. The calculations thus far have assumed con clearly indicates a flow direction toward the More likely sources for the alluvial fan are tinuous fault displacement between 3680 yr B.P. southwest. The source of the alluvial fan thus four closely spaced gullies several hundred and A.D. 1857. It is very likely, however, that must be on the opposite side of the San Andreas metres southeast of the fan apex (A, B, C, and D much, and possibly all, of the slip accumulates fault. Although the fan is composed of three in Fig. 1). In Figure 9, these have been restored sporadically, during large earthquakes, such as discrete beds in trench 2 (see detailed log of to their probable location at the time of forma that which occurred in 1857. If, as we argue trenches, available from author), the lack of tion of the fan. None of these four small gullies, below, this segment of the fault is characterized bioturbation or weathering of the two horizons which extend only a few hundred metres back by coseismic displacements of 10 m, followed between these beds suggests that the fan was from the fault scarp, could have been the sole by several centuries of quiet repose, the fault deposited very rapidly, perhaps in a matter of a provider of enough material to construct the en could have been at any point in its earthquake few decades or less. tire fan. The volumes of A, B, and C are only cycle 3680 yr B.P. If, in that year, the region The deposit overlies a massive, poorly sorted -13,000 m3 each, and D is much smaller. In bisected by the fault was in the middle or toward sandy loam that represents either a colluvial unit any combination, however, they could have the end of a period of elastic strain accumula or an alluvial deposit that was extensively bio delivered enough material. tion, the rate calculated using this data will be turbated prior to burial. The unit probably lay at The proper matching of this multiple source slightly too high, because the 128-m offset ac the ground surface for a long time prior to burial with the younger fan deposit can be determined cumulated between then and 1857 is in small by the alluvial fan. The presence of charcoal rather precisely. If the general reconstruction part due to loading that occurred slightly earlier. pebbles and granules in this unit, no more than a shown in Figure 9 is correct, the southeastern 66 SIEH AND JAHNS Figure 9. Isopach map of 13,250-yr-old alluvial fan and source gullies B and C. In this figure, the gullies have been restored 475 m to their late Pleistocene position upstream from the fan. The same gullies are indicated by letters B and C in Figure Ib. For reference, dotted line represents location of modern channel of Wallace Creek. Isopach map is based on trench exposures (thick, open bars) and geometry of contours. Insert in upper right illustrates use of topographic contours in constructing isopach map. Lower edge of stippled region is topographic contour. Upper edge is contour prior to deposition of fan. Southwestward bulging of contours indicates presence and thickness of alluvial fan. flank of the main fan complex had to be south D, only slightly less so than the oldest, beheaded not be distinguished from the average late Hrto- east of channel C. The offset thus is no less channel of Wallace Creek itself (marked with a cene rate of 33.9 ± 2.9 mm/yr. than 472 m. At the same time, the crests of the white arrow at the left margin of Fig. 1). The two distinct lobes of the fan shown in Figure 9 creation of gullies A, B, C, and D must, there RECURRENCE INTERVALS should have had their apexes at the mouths of fore, be nearly contemporaneous with the first BETWEEN PAST two of the middle gullies. Only gullies B and C entrenchment of Wallace Creek. LARGE EARTHQUAKES are spaced appropriately to meet this constraint. These considerations constrain the offset of AT WALLACE CREEK The mouth of gully B cannot be offset more than the younger fan deposits to 475 ± 3 m. In that 478 m, if B is the source of the northwestern the younger fan formed 13,250 ± 1650 yr B.P., The average Holocene and late Holocene lobe of the fan. It is of interest that the younger the average slip rate must be 35.8 + 5.4/-4.1 rates of slip at Wallace Creek are important new fan deposits are offset from gullies A, B, C, and mm/yr. Within the level of resolution, this can measurements of strain across the San Andreas 67 HOLOCENE ACTIVITY. SAN ANDREAS FAULT TABLE : SMALLEST STREAM OFFSETS NEAR WAl LACE CREFK TABLE 3 SMALLEST STREAM OFFSETS NtAR WALLACE CREEK AM> PROPOSED INTERVALS BFTWEFN GREAT EARTHQUAKES AND PROPOSED DATES AND CORRELATION OF LATEST FOUR GREAT EARTHQUAKES (1) (2l (-H (4) (51 Slip associated Proposed . III (21 (.11 (41 (Si n-am otfsi-i- with interval between Time required to Iml Remark^ Produced bs earthquake I m) events ( vri accumulate offset as Possible correlations Possible correlations clastic strain using average Proposed dates fur with events with events Stream late Holocene slip rate latest earthquakes recogm/cd at recogm/ed at Mill Average oflsets( mi ( years 1 (ADi Fallen Creek Potreroby Davis(l983l 5 mea :i« i ' Average of 1857 plus last 12.3 i : 1857 Z( 18571 ZU857) 4 measurements** prehistoric event 9.5-05 - 300 to 440* (-. lol 240 to 320 ?i 8 or 33.5 1 9 Average of 1 857 pi us latest 1 1 .0 or I540io 1630* V(I550 r 70) VI 1584 ;70| 3 measurements" 2 prehistoric events 11.7 i 2.2* 240 to 450* 21.8 I.I 560 to 740 Il20to 1300' RII080 - 65) 32.8 or 21.8 95 - (05*_j> I I") ' 33.5 - 1 9 840to 1140 *328 21.8 : (l.r * I9: )'"or335 21.8 r (I.I 2 » 1.92 )^ 720 to 1020' F(845 - 75) 'Slip dunng following earthquake in column 4 divided by average late Holocene slip rate (33.9 ± 2.9 mm/yr) "Offset gullies are all between Wallace Creek and Gully D in Figure I |857-(240to320yrl t|857-(560to740yr) §I857-(840to M40yr) fault in central California, because they are the first to span more than a fraction of a great earthquake cycle of strain accumulation and re lief. These millennial averages can be used in amount of fault slip associated with each of the Second, geodetic data on modern rates of conjunction with other data to infer earthquake last two prehistoric earthquakes. Table 2 lists the strain accumulation across the fault are available recurrence intervals. data that suggest "these 2 events were associated from the "Carrizo" net, which spans the fault For example, the length of the cycle of strain with -12.3 and 11.5 m of fault slip at Wallace and 80 km of adjacent territory at the latitude of accumulation that preceded and led to the great Creek. At 34 mm/yr, these values would have Wallace Creek (Savage, 1983, and 1982, writ 1857 earthquake can be calculated. In 1857, the accumulated in 360 and 340 yr, respectively. ten commun.). These data are available, how San Andreas fault sustained 9.5 m of right- The actual range in value for both of these recur ever, only for the period 1977.6 to 1981.5. The lateral slip at Wallace Creek. This is indicated rence intervals, calculated using the ranges in deformation observed during this period aver by five small offset gullies nearby (A, B, C, D, value for the slip rate and the offsets, is displayed ages 0.29 ± 0.06 MStrain/yr (extension) N89° ± and E in Fig. 1; Table 2), as well as by small in column 5 of Table 2. From the table, one can 4°W and -0.09 ± 0.06 jistrain/yr (contraction) offset gullies at distances of as much as several see that the latest 3 recurrence intervals are esti north-south. Numerous models of lithospheric kilometres to the northwest and southeast. These mated to be within the range of 240 to 450 yr. deformation can produce this observed surficial gullies were incised across the fault prior to the Of course, it is possible that the 4,000-yr and deformation. One class of model involves the 1857 event, but after the previous large event 13,000-yr average slip rates do not represent the assumption that the observed deformations are {see Sieh, 1978c, for a more detailed discussion). average rate of strain accumulation during the the result of aseismic right-lateral slip on the San If one assumes that the 9.5 m of fault slip asso periods of fault dormancy prior to 1857 and the Andreas fault beneath its locked, brittle upper ciated with the 1857 earthquake relieved elastic 2 previous great earthquakes. For example, the 10 or 20 km. In this case, the observed deforma strains that had accumulated in the adjacent rate of accumulation actually could have been tions of the Carrizo net are resolved as right- crustal blocks at an average rate of 34 mm/yr, much higher during the past millennium and lateral shear strains parallel to the San Andreas one calculates that the 1857 earthquake was much slower during the previous 4,000-yr inter fault. The average shear strain over the entire preceded by a 280-yr period of strain accumula val. If so, the recurrence intervals between the 80-km-wide network is 0.38 ± 0.04 ^rad/yr. tion. This calculation does not assume that a,n- latest few earthquakes would be much shorter This translates into a deep slip rate on the fault nual strain accumulation was uniform during than those calculated above. Perhaps a future of 30.4 ± 3.2 mm/yr, if one assumes that the the 280-yr period, but only that the average an study of a currently undiscovered 1,000-yr-old network spans the entire zone of deformation nual rate was equal to the millennial average of feature near Wallace Creek will resolve this due to slip on the fault. If it does not span the 34 mm/yr. Periods of faster or slower accumu issue by providing a 1,000-yr average rate. Al entire zone, the rate of deep slip on the fault lation thus could be accommodated within the ternatively, the past several earthquakes may be must be higher. Like the 13,000-yr average rate, over-all loading cycle. Table 2 (top of col. 5) dated directly, as has been done at Pallett Creek the geodetically determined modern rate does displays the actual range of values for the period (Sieh, 1978b, in press). In the meantime, the not differ significantly from the 3,700-yr of strain accumulation if the uncertainties in the validity of using the 3,700-yr average slip rate in average. 1857 offset value and average slip rate are taken calculating recurrence intervals of recent and fu The similarity of the 13,000-yr, 3,700-yr, and into account. In lieu of a direct dating of the ture great earthquakes must be assessed in other 4-yr averages suggests that strain accumulation large event that preceded the 1857 event at Wal ways. across the fault may be fairly uniform. Of lace Creek, this range (240-320 yr) is probably First, the slip rate averaged over the past course, numerous histories could be invented the best estimate that can be made for the recur 3,700 yr (33.9 ± 2.9 mm/yr) does not differ that include these three data points and yet in rence interval between the 1857 earthquake and appreciably from the 13,000-yr average (35.8 + volve large fluctuations in the strain accumula its predecessor. 5.4/-4.1 mm/yr), although the 13,000-yr aver tion rate between earthquake cycles or recur Estimates of the duration of two earlier peri age conceivably could be as much as 10 mm/yr rence intervals. To date, however, no known ods of strain accumulation can also be made, (-30%) faster than the late Holocene average, data support large fluctuations. A reasonable as using the average late Holocene slip rate and the given the imprecision of the 2 determinations. sumption, thus, is that the late Holocene average SIEH AND JAHNS 68 slip rate represents the average rate of strain ac phologic data, however, suggest that this is likely page during the past three to four large earth cumulation between large earthquakes. The re only along two portions of the 1857 rupture. quakes. Although, of course, so few data do not currence intervals displayed in Table 2 may. Wallace Creek is not within either of these provide a statistically sound basis for predicting therefore, be realistic estimates of the dormant portions. all previous and future events, we are confident intervals that preceded the past three great Figure 10 displays offsets measured along the that this pattern offers some insight into the earthquakes. south-central segment of the San Andreas fault. long-term behavior of the fault. In the next section, we attempt to assess when The 1857 segment is divisible into at least three At least two explanations are worth consider the current earthquake cycle will end at Wallace parts, based on slippage during the 1857 earth ing. First, we consider the possibility that the Creek; that is, when the next great earthquake, quake and one to four previous large earth northwestern 40 km and southeastern 90 km are accompanied by rupture at Wallace Creek, will quakes. The southeastern part is -90 km long loaded more slowly, and, therefore, when the occur. We also attempt to use the 3,700-yr aver and seems to have been characterized by 3- to earthquake occurs, they experience lesser age slip rate to assess the likelihood of large 4.5-m slip events. The central 160 km, including amounts of slip than does the central 160-km- earthquakes elsewhere along the San Andreas Wallace Creek, has experienced 7 to 12.3 m of long part. This is unlikely, because the average fault. slip during the most recent 3 great earthquakes. Holocene slip rate along these two parts must be The lower values along this central portion nearly equal to the rate determined at Wallace FORECASTS OF THE BEHAVIOR occur along the reach between km 90 and km Creek. Just beyond the south-central segment, at OF THE SAN ANDREAS FAULT 200, where several other active faults to the Cajon Creek (Fig. 10), the San Andreas has av north and northeast exist, and so the lower erage Holocene and late Holocene slip rates of Along the South-Central (1857) Segment values may reflect distributed deformation, 25 ± 3 mm/yr (Weldon and Sieh, 1981). The away from the San Andreas fault. A 30-km nearby San Jacinto and related subparallel faults If the crust adjacent to the San Andreas fault segment northwest of Wallace Creek expe probably carry -10 mm/yr at this latitude has been accumulating strain at 34 mm/yr since rienced 3 to 4 m of slip in 1857 and probably 1 (based on data of Sharp, 1981, and Metzger, 1857, as much as 4.3 m of potential slip has now or 2 during previous large earthquakes, as well. 1982). These fault systems end and nearly merge been stored and conceivably could be released These data suggest that each part of the fault with the San Andreas fault just northwest of along all or part of the 1857 rupture. Geomor- has experienced a characteristic amount of slip- Cajon Creek. Farther northwest; the San An dreas fault is the only major active structure, and so northwest of Cajon Creek, it must have a slip rate of -35 mm/yr. In addition, the average NW OKM 200 300 SE recurrence interval for large earthquakes at Pal- 60 leu Creek (location in Fig. 10) is in the range of 145 to 200 yr, which is appreciably shorter than the 240- to 450-yr range at Wallace Creek. For 50 this reason, some of the slip events shown in Figure 10 in the Palmdale-Pallett Creek region must have their northwestern rupture tip south east of Wallace Creek, and 1857-like events cannot be the only type of slip event along this o part of the south-central segment The northwestern 30 km of the south-central « 30 segment (Fig. 10) must also share the long-term average slip rate of Wallace Creek. No diversion of a large fraction of the Wallace Creek rate r-o O along other structures is plausible. The only known major active(?) fault nearby is the San Juan Hill fault, which runs 3 to 14 km west of " 10 and subparallel to the San Andreas from about Cholame to Wallace Creek (Jennings and oth ers, 1975). Its rate of slip is probably no more than a few millimetres per year. A second explanation for the different behav ior of the three parts of the south-central seg ment is based on the hypothesis that each part is imbued with a different strength. If, for reasons Figure 10. Right-lateral offsets measured along the south-central (1857) segment of the San of geometry or rock properties, the central 160 Andreas Fault suggest that slip at each locality is characterized by a particular value. Solid km of the segment were 2 or 3 times stronger circles are data from Sieh (1978c), with poor-quality data deleted. Open circles are data From than the 2 other parts, 2 or 3 times as much Davis (1983). Triangles are new data and remeasurements at sites reported by Sieh (1978c). elastic loading of the adjacent crustal blocks Open squares are new data. Vertical bars indicate magnitude of imprecision in measurement. would be necessary before failure occurred. 69 HOLOCENE ACTIVITY, SAN ANDREAS FAULT Each failure thus would result in two to three terey Bay (Savage, 1983; Lisowski and Prescott, upon to absorb a large portion of the slip rate times as much slippage as on the two adjacent 1981). The long-term rates at Wallace Creek are observed farther south at Wallace Creek. Like parts. Such an explanation is compatible with also identical to the historical rate of slip at shal wise, there are no obvious geological structures our judgments that (1) slip rate does not vary low levels along the central 50 km of the creep near the San Andreas that would lead one to crcatly along the south-central segment, and ing segment (see data compiled by Lisowski and suspect that the long-term slip rate along the (2) large earthquakes are more frequent at Prescott, 1981, Fig. 6). These similarities could creeping segment is appreciably higher than the Pallet! Creek than at Wallace Creek. be coincidental, but they suggest that the central long-term rate farther south. Table 3 lists our best estimates of the dates of 50 km of the creeping segment is creeping annu Along the 90-km Segment large earthquakes-at Wallace Creek and pro ally at its millennial-average rate of slip. If this Centered on Cholame posed correlations with large earthquakes that were true, it would mean that large elastic have been directly dated at Pallett Creek (Sieh, strains are not accumulating across the central Between the central 50 km of the creeping in press) and at Mill Potrero (Davis, 1983). The 50 km of the creeping segment, and that this zone and Wallace Creek, there is a stretch of the capital letters in Figure 10 reflect our best judg segment will not participate in the generation of San Andreas fault that historically has been a ment regarding correlation of the latest events at the next large earthquakes along the San An zone of transition between the fully creeping and Wallace Creek. Pallett Creek, and Mill Potrero. dreas fault. fully locked portions of the fault. On the basis of Event X at Pallett Creek (A.D. 1720 ± 50) has From a geological point of view, it is reason available data, this segment is a prime candidate no correlative at Wallace Creek, although Davis able to suspect that the long-term slip rate along for generating a large earthquake in the near (1983) discovered evidence for and dated a rela the San Andreas fault at Wallace Creek should future. In the period of historical record, it has tively small slip event at Mill Potrero that may not be different from its long-term rate along the not experienced as much slip as have segments well be event X. Event V at Pallett Creek oc creeping segment, except along its northernmost to the northwest or southeast, and it is therefore curred about A.D. 1550, which is about the time 50 km, adjacent to which runs the actively a "slip gap." we estimate that the last prehistoric event at creeping Paicines fault (Harsh and Pavoni, One interpretation of the historical data is il Wallace Creek occurred, and also about the 1978). No other large, active structures in the lustrated in Figure 11, in which cumulative time of a large slip event that Davis (1983) dis latitudes of the creeping segment can be called right-lateral slip for the past two centuries is covered at Mill Potrero. Similarly, events R and F at Pallett Creek occurred at about the time we estimate that the third and fourth events oc curred at Wallace Creek. On the basis of the foregoing discussion, we judge that the central 160 km of the south- central segment of the San Andreas fault is un likely to generate a great earthquake for at least another 100 yr. Recurrence intervals appear to be in the range of 250 to 450 yr, and yet the time elapsed since the great earthquake of 1857 is only 127 yr. Slip during the latest 3 great earth quakes has been 7 to 12.3 m, and yet we suspect that only a little more than 4 m of potential slip has been stored in the past 127 yr. The southeastern 90 km and the northwestern 30 km of the south-central segment are good candidates for producing a large earthquake within the next several decades. Geomorpho- logic measurements seem to indicate that 3 to 4.5 m of slip is characteristic during large events, and >4 m of potential slip may well have been stored in the adjacent crustal blocks since 1857. Based on studies at Pallett Creek, the probability of a great event along the southeastern 90 km of the south-central segment within the next 50 yr Figure 11. Hypothetical source of future major earthquake along the San Andreas fault is between 26% and 98% (Sieh, in press). includes 60 km of the currently creeping segment and 30 km of the locked segment. Cumula tive right-lateral slip plotted against distance along the fault indicates that this 90-km segment is Along the Creeping Segment slip-deficient relative to adjacent stretches of the fault. Slip in 1857 is from Sieh (1978c). Cumulative slip along the creeping segment is extrapolated from alignment array slip rates for The long-term average slip rates determined period 1968-1979 (Lisowski and Prescott, 1981, Fig. 6). Dates of moderate earthquakes at Wallace Creek are indistinguishable from the generated by slip along the fault in the Parkfleld-Cholame region are shown, because such an geodetically determined rates of slip at deep lev event probably triggered the great 1857 rupture and conceivably could trigger the rupture of els along the fault from Wallace Creek to Mon- the slip gap. SIEH AND JAHNS 70 plotted as a function of location along the fault. ROLE OF THE SAN ANDREAS REFERENCES CITED We assume that creep rates northwest of Cho- FAULT IN THE RELATIVE MOTION Ajincvs. I) and Sieh. K . |s)7N. A documentary Mud> <>l the felt effects ol the great California earthquake of 18V SasmologiLal Socictv of Amenta lame have been constant for the past few OF THE NORTH AMERICAN Bulletin, i 6X. p 1717 1729 Arnold. R . and Johnson. H R . 1909. The earthquake rift in eastern San l.uis hundred years, so that the alignment array data AND PACIFIC PLATES Ohispo Counts, (ahlorma Science. v 29.no 744. p SSX for the period 1968-1979 are representative of Cm*ell. J . I"")- Displacement along the San Andreas f-ault. California Geological Socieu of America Special Papers. s 71. 61 p the pre- and post-1857 creep rates. We also as Minster and Jordan (1978) determined from 19X1. An outline of the ieuom<. historv of southeastern California, in trnst. W (j. ed The gcoteuonu. development of California Engle- sume that Cholame has been the edge of the a circumglobal data set that the relative motion wmid Cliffs. New Jcrscv. Prentice Hall, p 5X4 MX) of the Pacific and North American plates has Da\is. T . 19X3. Late Cenopoic structure and tectonic history of the western creep zone throughout this period. In the cen -BiB Bend" of the San Andreas Fault and adjacent San Emigdio Moun tury preceding-1857, 3 to 3.5 m of slip would averaged -56 mm/yr during the past 3 m.y. tains [Ph D dissert ] Sania Barbara. California. University of Califor nia. Department of Geological Sciences have accumulated by creep northwest of Slack The geological record at Wallace Creek shows Harsh. P. and Pavom. N. 1978. Slip on the Paicmes fault Seismological Societ> of Amenca Bulletin, v 68. p 1191 1194 Canyon. Less slip would have accumulated by that, at least during the past 13,000 yr, only -34 Hill. M . and Dibblee. T . Jr. 1953. San Andreas. Oarlock and Big Pine faults. creep, and perhaps during occasional moderate mm/yr of this has been accommodated by slip California A studs of their character, hisiorv and tectonic significance of their displacements Geological Societ> of Amenca Bulletin, v 64. earthquakes, between Slack Canyon and along the San Andreas fault. If one assumes that p 44.1 45K Jenmngs. C . and other.. 1975. Fault map of California California Division of Cholame. the 3-m.y. average represents the Holocene av Mines and (icologv California Geological Data Map Series Map no I In 1857. -3.5 m of slip occurred along the erage rate across the plate boundary as well, Klein. J lerman J C Damon. P E . and Ralph. E K . 19X2. Calibration of radiocarbon dates Tables based on the consensus data of the Workshop 30-km stretch of the fault southeast of Cholame, then clearly the San Andreas fault is accommo Calibrating the Radiocarbon Time Scale Radiocarbon. v 24. p 10.1 ISO and 9.5 m of slip occurred in the vicinity of dating only -60% of the relative plate motion. Lawson. A. and others. I90X The California earthquake of April 18. Wallace Creek. The sparse historical accounts The remainder of the deformation must be ac 1906 Report of the State Earthquake Investigation Commission Wa shington. DC. Carnegie Institution of Washington. 2 volumes and are compatible with our inference in Figure 11 complished elsewhere within a broader plate atlas. 461 p Lisowski. M . and Prescott. W H.. 1981. Short range distance measurements that slippage during the earthquake decreased boundary. The San Gregorio-Hosgri fault sys along the San Andreas fault system in central California Seismological tem, which traverses the coast of central Cali Socieu of Amenca Bulletin, v 7l.no 5. p 1607 1624 northwestward from Cholame and died out near Metzger. I.. 1982. Tectonic implications of the Quaternary history of Lower Slack Canyon (Sieh, 1978c, p. 1423-1424). fornia, may have a late Pleistocene-Holocene Lvtle Creek, southeastern San Gabriel Mountains |B.A thesis] Clare- mom. California. Pomona College Following 1857, creep resumed northwest of slip rate of 6 to 13 mm/yr (Weber and Lajoie, Minster. J B. and Jordan. T. H.. 1978. Present-day plate motions Journal of Geophysical Research, v. 83. no Bl I. p. 5331 5334 Cholame. Northwest of Slack Canyon. 4.5 m 1977). and the Basin Ranges, to the east of the Ntlsen. T.. and Link. M H.. 1975. Stratigraphy, sedimentologv and offset along of slip now has accumulated at the full, long- San Andreas fault, may be opening N35°W on the San Andreas fault of Eocene to lower Miocene strata of the northern Santa Lucia Range and the San Emigdio Mountains. Coast Ranges, term rate of loading of the fault (that is. 34 oblique normal faults at a late Pleistocene-Hol central California, in Weaver. D. W . and others, eds.. Paleogene Sy m- posium and selected technical papers Conlerence on Future Energy mm/yr). The 60-km-long section between Slack ocene rate of ~7 mm/yr (Thompson and Burke, Horizons of the Pacific Coast, Annual Meeting AAPG-SEPM-SEG. Canyon and Cholame, however, has crept at 1973). Most of the 56 mm/yr plate rate thus Long Beach. California, p 367 400 Reid. H f. 1910. Permanent displacements of the ground, in The California rates that are significantly lower than the loading may be attributed to the San Andreas, San Gre earthquake of April 18. 1906- Report of the State Earthquake Investi gation Committee: Washington. DC.. Carnegie Institution of Washing- rate, and strain is being stored in the rocks adja gorio-Hosgri, and Basin Range faults. Long- ion, v 2. p. 16-28 cent to the fault there. Similarly, elastic strains term slip rates on these three major fault systems Savage. J. C. 1983. Strain accumulation in western United Stain. Annual Reviews of Earth and Planetary Science, v 11. p M - 43 are accumulating in the rocks adjacent to the are not known precisely enough to preclude or Sharp. R. V. 1981. Variable ram of late Quaternary strike slip on the San Jacinlo fault zone, southern California: Journal of Geophysical Re locked portion of the fault, and the northern confirm the possibility that the rate of relative search, v 86. p. 1754-1762 plate motion during the Holocene is equal to the Sieh. K.. 1977. Late Holocene displacement history along the south-central most 30 km of this portion, which seems to fail reach of the San Andreas Fault (Ph D dissert.) Stanford. California. in 3- to 4-m slip events, may well be loaded 3-m.y. average. No clear basis exists, however, Stanford University. 219 p. I978a. Central California foreshocks of the great 1857 earthquake: nearly to the point of failure. for suggesting that the Holocene rate is less than Seismologtcal Society of America Bulletin, v 68. p. 1731-1749 I978b. Pre-hisionc large earthquakes produced by slip on the San We suggest that this northernmost part of the or more than the longer-term rate. Andreas fauli ai Fallen Creek. California: Journal of Geophysical Re locked segment and the southernmost part of the search, v 83, p. 3907-3939. I978c. Slip along the San Andreas fault associated with the great 1857 creeping segment might fail in unison and pro earthquake: Seismological Society of America Bulletin, v. 68. p. 1421-1428. duce a major earthquake. This hypothetical in press. Lateral offsets and revised dates of large prehistoric earthquakes event would be associated with -90 km of sur at Pallet! Creek, southern California. Journal of Geophysical Research Stuiver. M.. 1982. A high-precision calibration of the AD radiocarbon time face rupture and a maximum of 4.3 m of right- ACKNOWLEDGMENTS scale Radiocarbon, v. 24. no. I. p. 1-26. Sluiver. M.. and Polach. H. A. 1977. Discussion Reporting of I4C data. lateral slip. Radiocarbon, v. 19, no. 3. p 355 363 In discussing this hypothetical event, it is im Wallace Creek is named after Robert Wal Thaicher. W . 1975. Strain accumulation on the northern San Andrcas fault zone since 1906 Journal of Geophysical Research, v 80. no 35. portant to note that the great 1857 earthquake lace, who elucidated the basic history of the p. 4873 4880 Thompson. G. A., and Burke. D. B.. 1973. Rate and direction of spreading in seems to have originated in this region. Sieh channel more than 15 years ago and provided us Dixie Valley. Basin and Range province. Nevada: Geological Society of (1978a) documented that at least 2 moderate with a special topographic base map. N. Amenca Bulletin, v. 84. p. 627-632 Wallace. R. E.. 1968. Notes on stream channels offset by the San Andreas fault. , foreshocks occurred in this vicinity about 1.5 Timothy Hall drew our attention to the study southern Coast Ranges. California, in Dickmson. W . and Grantz. A.. eds.. Conference on Geologic Problems of San Andreas Fault S^tem. and 2.5 hr prior to the main shock. Within the site. He and Laurie Sieh participated in initial Proceedings Stanford University Publications in the Geological Sciences, v. 11. p. 6-21. past century, 5 moderate (M5.5 to 6) earth studies. Art Fairfall and John Erickson at the Weber. G E.. and Lajoie. K R.. 1977. Late Pleistocene and Holocene tectonics quakes have been produced by slippage along University of Washington provided all of the of the San Gregono fault zone between Moss Beach and Point A no . Nuevo. San Mateo County. California Geological Society of Amenca the San Andreas fault northwest of Cholame. radiocarbon analyses. Robert Wallace, David Abstracts with Programs, v 9, no 4. p. 524. Weldon. R. J . and Sieh. K. E.. 1981. Offset rate and possible timing of rec*nt Sieh (1978a) inferred that the 1857 foreshocks Schwartz, David Pollard, Christopher Sanders, eanhquakes on the San Andreas fault in Cajon Pass. California (abs.| emanated from a source similar to that which and Ray Weldon provided helpful criticisms of EOS (American Geophysical Union Transactions), v 62. no 45. p. 1048. produced these historical shocks. If this is true, earlier manuscripts. This work was supported by MANI* DIPT RFC FIVFOav THFSlXIFTV NoVfSIBFR 10. 1982 then the next moderate "Parkfield-Cholame" the National Earthquake Hazards Reduction Rf.usn>MAs.L*HinRFcFivFt>SFrminn<2. 1983 Program, U.S. Geological Survey Contract nos. Nl4M.scRIPI A(CFPTfnSFCTFMHiR22. 1983 earthquake might well be a foreshock of the CONTRIHLTION No 3819. DIVISION OF GFOLOGIOAL AND PLAMTART hypothetical major event described above. 14-08-0001-15225, 16774, 18385, and 19756. ScifNCFi CALIFORNIA INSTITUTE OF TFCHNOLOOV Printed in U.S.A. 71 APPENDIX C Terms for Expressing Earthauake Potential, Prediction, and Probability Robert E. Wallace, James F. Davis, and Karen C. McNally reprinted from the Bulletin of the Seismological Society of America, vol.74, no. 5, with permission of the Seismological Society of America 72 Bulletin of th« Sei«roolog>c«l Society of America. VoL 74, No. 5, pp. 181&-1B25, October 1984 TERMS FOR EXPRESSING EARTHQUAKE POTENTIAL, PREDICTION, AND PROBABILITY BY ROBERT E. WALLACE, JAMES F. DAVIS, AND KAREN C. MCNALLY ABSTRACT Terms for expressing earthquake potential and prediction include two main categories, "long-term earthquake potential" and "earthquake prediction." Earth quake prediction is subdivided into three categories "long-term prediction," "intermediate-term prediction," and "short-term prediction." Long-tarm prediction Is not subdivided, but two terms, "watch" and "forecast" are recognized as having similar meanings. "Short-term prediction" is subdivided into "alert" and "imminent alert" The subdivisions of earthquake prediction are based on differ ent time frames. Earthquake potential or probability can be expressed either numerically or verbally according to a variety of schemes. . 1 . ..-_ ___._. INTRODUCTION AND HISTORY In response to the Earthquake Hazards Reduction Act of 1977, a report, published in 1978 under the auspices of the Office of Science and Technology Policy (Working Group on Earthquake Hazard Reduction, 1978), was concerned with "issues fot an implementation plan." One of the issues cited was the need for standardization of terms such as "prediction," "alert," and "warning." In 1980 the Southern California Earthquake Preparedness Project (SCEPP) was begun under the auspices of the California Seismic Safety Commission and the Federal Emergency Management Agency, and a similar need for standardized terms for emergency service and public . response planning was recognized. The present authors were designated as a committee of the Policy Advisory Board of SCEPP to consider predictive terms and their application. An early version of the terminology, which included a probability element as j^art of the definitions, was tentatively adopted by SCEPP in December 1981 and by the California Earthquake Prediction Evaluation Council in April 1982. The version described in this report, which excludes probability from the definitions, has been incorporated into some SCEPP planning documents of 1983 and was formally approved by the Policy Advisory Board of SCEPP on 28 September 1983 and by the California Seismic Safety Commission on 13 October 1983. The earlier version of proposed terminology was reviewed by the National Earthquake Prediction Evaluation Council in June 1982, but action was postponed pending further study and possible revisions. The present version reflects some concerns and suggestions of that panel In a report by the National Academy of Science-National Research Council (U.S. National Research Council, Panel on Earthquake Prediction of the Committee on Seismology, 1976), earthquake prediction was defined as follows: "An earthquake prediction must specify the expected magnitude range, the geographical area within which it will occur, and the time interval within which it will happen with sufficient precision so that the ultimate success or failure of the prediction can readily be judged. Moreover, scientists should also assign a confidence level to each prediction." In other documents, a distinction between "prediction" and "warning" is recom mended (Panel on Public Policy Implications of Earthquake Prediction, U.S. National Research Council, 1975; McKelvey, 1975). Both define "prediction" much 73 ROBERT E. WALLACE, JAMES F. DAVIS, AND- KAREN C. MCNALLY as does the National Research Council, but, according to McKelvey, "warning is a recommendation or an order to take some defensive action," and, according to The Panel on Public Policy Implications of Earthquake Prediction "warning is a decla ration that normal life routines should be revised for a time." Interpretation of the meaning of the term "warning" has not remained constant; e.g., in the following paragraph, the term "warning" has been defined as similar to "prediction" above. Under the Disaster Relief Act of 1974, the Director, U.S. Geological Survey (USGS) was given, by redelegation, the responsibility for issuing information about geologic hazards. After that time, terms for use in issuing warnings were published in the Federal Register (1977), and the following three terms were used by the USGS until October 1983 for official releases for all geologic hazards including earthquakes, volcanic eruptions, and landslides. Notice of potential hazard The transmission to Federal, State, and local officials of information about the location and nature of potentially hazardous geologic conditions. Evidence is insufficient to suggest that a hazardous event is imminent or to determine the time of occurrence. Hazard watch The transmission of information that a change is taking place in a geologically hazardous situation that may be interpreted as precursor to a potentially hazardous event within an unspecified period of time. Hazard warning The transmission of information about precursory phenomena that appear to signal a potentially hazardous event within a specific period of time (possibly days or hours). Official "earthquake watches" were issued by the Director, USGS for the southern part of the San Andreas fault in California in 1976 and for the Mammoth Lakes area of eastern California in 1980. - - The three-level classification, "notice," "watch," and "warning," although useful, seemed to confuse some public officials as well as the media and general public. A notice published in the Federal Register on 11 October 1983 (Devine, 1983) proposed that the USGS should henceforth use the term "Geologic Hazard Warning." The type of geologic hazard and its characteristics, such as area affected and imminence, are dealt with in a supplementary text. Terminology needs appear to be somewhat different for different geologic hazards. Long-term predictions of volcanic eruptions, e.g., are viewed in a time frame of weeks (Swanson et aL, 1983), whereas long-term predictions of earthquakes are viewed in the time frame of decades. The National Weather Service (NWS) has developed a comprehensive set of weather-related terms, and although some of the principles used in that terminology can be transferred to earthquakes, specific terms cannot. The use of adjective modifiers as employed by NWS, such as in "traveler's advisory" or "stockman's advisory," is a self-explanatory way of creating terms more specific than implied by the base term and is highly recommended. Despite formalization of terms to be used for any discipline or problem, the news media seldom abide by the technical definitions of terms. For example, the terms "notice," "watch," and "warning" were reported in newspapers as meaning "alert stage one, two, and three." Air pollution also is reported commonly in southern California as "alert stages one, two, or three" even though the term "air pollution episodes" is used officially (South Coast Air Quality Management District, 1981). Management District, 1981). Standardization of terminology is necessary for accurate and simple communi cation. But three principal audiences for earthquake-prediction terminology have rather diverse needs and different capabilities of understanding terms. The scientist 74 TERMS FOR EARTHQUAKE POTENTIAL, PREDICTION AND PROBABILITY must communicate with other scientists, the scientific community must communi cate with the disaster-response administrative community, and these two commu nities must communicate with the public. The terms suggested in this paper are aimed primarily toward the chain of communication, scientist to administrator to the public, but we have also considered what the scientist-to-scientist link may need (e.g., Aki, 1980). Usage will ultimately determine the success of any term, and so it will be with the terms suggested here. TERMS SUGGESTED We adopt and recommend use of the basic definition of earthquake prediction as presented by the U.S. National Research Council, Panel on Earthquake Prediction of the Committee on Seismology (1976), and here focus on the time element of a prediction as a needed refinement of the basic definition. Other elements of a prediction, i.e., place, size of earthquake, and likelihood of an event occurring as predicted are not a part of the defined terms except for the interdependence, to some extent, of statements of time and magnitude, a relation discussed later. The place or area in which an earthquake may occur are handled in a descriptive way and, for one example, are represented by maps of active faults. The size of the predicted earthquake can be expressed as Richter magnitude, seismic moment, or moment magnitude, or other commonly used expressions of earthquake size. The likelihood of an earthquake occurring as predicted can be stated in mathematical probability terms or by percentage chances per unit of time or in verbal form. The terminology framework suggested includes a ranking system similar to the family-genus-species ranking of biologic taxonomy. Thus, the terms "prediction" and "long-term earrthquake potential" are the highest of three ranks. The terms "short-, intermediate-, and long-term predictions" are second highest rank. Earth quake "alert" is of the lowest rank. As earthquake-prediction science improves, new subdivisions of each may be warranted and can be accommodated, we hope, without restructuring the overall framework. Use of the term "time window" (see below) may carry two connotations: (1) that the earthquake can occur at any time from the present through the period of the time window, or (2) that the period of the time window will pass before the earthquake is likely to occur. We suggest that for the present state of earthquake- prediction science connotation (1) will be the most useful, and should constitute the meaning unless otherwise indicated, but as the state of prediction science advances, connotation (2) may become needed. At such a time, subdivisions of long-, intermediate-, and short-term prediction can be created and used to distin guish the two meanings, or the specific meaning can be described. SUGGESTED TERMS Time Window Long-Term Earthquake Potential No specific time window. Can refer to decades, centuries, or millennia. Earthquake Prediction Any specific time window shorter than a few decades. Long-term prediction Few years to a few decades. Intermediate-term prediction Few weeks to few years. Short-term prediction Up to a few weeks. 75 ROBERT E. WALLACE, JAMES F. DAVIS, AND KAREN C. MCNALLY DEFINITIONS AND DISCUSSION Long-term earthquake potential. The potential or probability of an earthquake occurring in a given area or region, or on a given fault, can be expressed, e.g., in percentage chance per year or average recurrence interval for earthquakes of designated magnitude levels. No specific period of time of occurrence and no specific future earthquake is designated. The potential may remain the same for long periods of time, even hundreds of years. Discussion. Statements of "long-term potential" relate to probability based either on the historical or geologic record, or both. The long-term potential can be stated as average recurrence interval or percentage chance per year at designated magni tude levels. For very active faults, such as major elements of the San Andreas fault system, the values may fall in the range of 0.2 to 2 per cent per year for M 1 or greater earthquakes. For most other faults in the Western United States, the value is less than 0.2 per cent per year for major earthquakes. For example, earthquakes of M 7.5 to 8 on the northern San Andreas and Hayward faults are assigned a probability of 1 per cent chance per year, and an earthquake of M 7.5 on the Newport-Ingle wood fault system is assigned a probability of less than 0.1 per cent chance per year (Federal Emergency Management Agency, 1981; Lindh, 19S3). Most of the faults in the Basin and Range province are assigned a probability of less than 0.1 per cent chance per year for generating earthquakes greater than M 7 (Wallace, 1981). Earthquake prediction. An earthquake prediction specifies the expected magnitude range, the geographical area within which a specific future earthquake will occur, and the time interval within which it will happen. A confidence level is included in each prediction. Note that this definition is the same as that suggested by the U.S.- National Research Council, Panel on Earthquake Prediction of the Committee on Seismology (1976), but emphasizing a "specific future earthquake" helps to distin guish a prediction from a statement of long-term potential. Discussion. A prediction designates a specific period of time for the occurrence of a specific future earthquake of a given magnitude range; in contrast, statements of earthquake potentials apply for indefinite periods of time. The distinction between potential and prediction may not always be clear, and to some extent the designation is optional with the predictor. For example, if a statement were made that a 0.1 per cent probability per year exists, few would interpret such a statement to mean that an earthquake is predicted to occur within 1 yr. If, on the other hand, the probability was 50 per cent or greater per year, the statement would be interpreted by most as a "prediction" of an earthquake occurring in 1 yr. In general, we believe that if a probability greater than 50 per cent is expressed for any specific period of time, the public will consider the statement to be a prediction for that period. Probability values increase as the length of the time window increases, or the stated magnitude decreases, thus large probabilities can be stated if the time window is made long enough or the magnitude of the expected earthquake is made small enough. Conceivably, however, premonitory evidence may suggest a very specific time in the future for an event of a given size. Furthermore, a situation can be imagined in which data would indicate a specific time and magnitude, but the confidence level would be very low. In such cases, an interchange between proba bility, magnitude, and length of time window would not be possible. The subcategories of "prediction," i.e., "long-term prediction," "intermediate- term prediction," and "short-term prediction," are based only on length of time 76 TERMS FOR EARTHQUAKE POTENTIAL, PREDICTION AND PROBABILITY windows, even though, as stated above, an interdependence of time frame, magni tude, and probability exists. Two possible meanings of "time window" are discussed above, but we suggest that the science should be permitted to evolve further before specific terms are adopted to include these differences in a formal taxonomy. Long-term prediction. A prediction of an earthquake that is expected to occur within a few years up to a few decades. Discussion, The terms "forecast" and "watch" have been used previously and carry connotations similar to long-term prediction. The term "forecast" is a state ment of future expectation and, for the present, may be used synonymously with long-term prediction. To some, the term "forecast" connotes less specificity than prediction, but the distinction is moot. The term "watch" carries the connotation of continuous attention to the situation, possibly including increased monitoring of an area. Intermediate-term prediction. A prediction of an earthquake that is expected to occur within a period of a few weeks to a few years. No subdivisions are suggested now, Short-term prediction. A prediction of an earthquake that is expected to -occur within a few hours to a few weeks. Discussion. The terms "alert" and "imminent alert" may be used as subdivisions of short-term predictions. The term "alert" carries a sense of urgency, e.g., Webster defines "alert" as "an alarm or other signal to warn of danger." The news media commonly have used the term "alert" regardless of what terms are used by public officials or scientists in formal notices or advisories. The term "alert" is applied to the period of 3 days to a few weeks. The term "imminent alert" is applied to a period up to 3 days. Advice from disaster-response administrators suggests that maintenance of the highest level of readiness beyond 3 days would be difficult. Because the terms "alert" and "imminent alert" convey a sense of urgency, we suggest that these subdivisions or "species" of short-term prediction be used only when the probability level is high. If the probability level is low or moderate, use of only the generic form "short-term prediction" is recommended. OTHER TERMS Numerous other terms are useful and, to the extent that they convey general meaning, may be employed. We caution against restricting the definitions of any word or term in such a way that the restricted definition carries the meaning outside the bounds of the generally understood dictionary meaning or excludes elements of the generally understood dictionary meaning. Misinterpretations are too easily drawn. Advisory A formal message giving earthquake information or advise to take action. Area of intensified study Either a formal or informal recognition of special interest, study, or monitoring in an area, because of an increase in the perceived likelihood of an impending earthquake (see Japanese National Land Policy Series, Law No. 73). Intensified study of an area may be initiated for scientific purposes, such as to conduct a prediction experiment, or because the area is densely populated and the hazard potential is high. Notice The formal communication of earthquake information, especially earth quake-potential information. Tendency A term- implying a dynamic, changing situation as contrasted to the 77 1 ROBERT E. WALLACE, JAMES P. DAV1S, AND KAREN C. MCNALLY static long-term potential. It relates to prediction and is used by the Chinese, who hold National Earthquake Tendency Conferences. The term indicates that a phys ical process is under way that may lead toward an eventual earthquake. The "tendency" can be "weak," "moderate," or "strong" and can "increase" or "decrease." Warning The generally understood meaning of "warning," as defined by Webs ter's dictionary, is "the action or fact of putting one on his guard by intimating danger." Various restricted definitions of the term "warning" have been proposed and used in connection with earthquakes as well as other natural hazards such as severe weather. The USGS defines "'geologic hazard warning' as a formal statement by the Director of the U.S. Geological Survey that discusses a specific geologic condition, process or potential event that poses a significant threat to the public, and for which some timely response would be expected" (Devine, 1983). .. - . _.. ..____. pOTENTIAL AND PROBABILITY ------Potential and probability can be expressed in various ways, and the form will depend to a large extent upon the audience. The scientific community may have a different perception of the significance of hazard probability than the public. For example, a probability gain from 0.009 to 0.09 means an order of magnitude increase to the scientist and yet the 0.09 translated into 9 per cent chance could be considered a low probability by the public. Furthermore, the probability of a geological event (hazard) occurring compared to the probability of a risk (lives or dollar value) are different aspects. Further study of these problems is needed. Long-term potentials have been expressed as percentage chance per year for earthquakes of a given magnitude, as average recurrence intervals on individual - faults, as recurrence intervals normalized for areas, in map form showing expected accelerations and velocities of ground motion region by region, as well as other ways. For long-term predictions, probabilities have been expressed for a specific future event as probability (in a mathematical notation of 1 equal to 100 per cent chance) per day (Aki, 1981), percentage chance per year, and percentage chance per 30 yr_ _ (Lindh, 1983). As an example, Lindh estimates that the Parkfield segment of the central San Andreas fault has a 99 per cent chance of generating an M 6 earthquake in the next 30 yr. According to the suggested terminology, such a statement is a "long-term prediction." We have been convinced by discussion with members of the news media that, in general, expressions of percentage are not well understood by all members of the lay audience. We suggest, therefore, that tne following words be used as equivalent to a three-fold division of percentage values regardless of time period 0-10 per cent Slight, as "slight chance of an earthquake" 11-49 per cent Moderate 50-100 per cent High . Some members of the lay public appear to understand percentage better if stated as "one chance in ten" or "seven chances in ten," as chances of rain commonly are stated. TERMINATION OR REDUCTION OF ANY LEVEL Any of the levels of prediction can be terminated, reduced, or modified at any time as the geophysical or other anomalies or interpretation of anomalies change. 78 TERMS FOR EARTHQUAKE POTENTIAL, PREDICTION AND PROBABILITY RESPONSIBILITY OF PREDICTORS At the meeting of the Board of Directors of the Seismological Society of America in April 1983, a set of "Guidelines for Earthquake Predictors" was approved by the Board and these guidelines are published in the Bulletin of the Seismological Society of America (1983). We urge all scientists engaged in formulating or issuing earthquake predictions to familiarize themselves with the problems considered in those guidelines. ACKNOWLEDGMENTS Of the numerous people who have commented on the problems of terminology, we especially wish to acknowledge members of the staff and Policy Advisory Board of the Southern California Earthquake Preparedness Project, members of the National Earthquake Prediction Evaluation Council, the California Earthquake Prediction Evaluation Council, and participants in the 1981 and 1984 workshops of the Southern California Earthquake Preparedness Project. - - REFERENCES Aki, K. (1981). A probabilistic synthesis of precursory phenomena, in Earthquake Prediction: An International. Review, D. W. Simpson and P. G. Richards, Editors, Am. Geophys. Union, Maurice Ewing Ser. 4, 566-574. Devine, J. F. (1983). Revision of terminology for geologic hazard warnings, Federal Register 48, 46104- 46105. Federal Emergency Management Agency (1981). An Assessment of the Consequences and Preparations for a Catastrophic California Earthquake: Findings and Actions Taken: Washington, D.C. 59 pp. Londh, A, G. (1983). Preliminary assessment of long-term probabilities for large earthquakes along selected fault segments of the San Andreas fault system in California, UJS. GeoL Sun., Open-File Rept. 83-63,15 pp. McKelvey, V. E. (1975). A federal plan for the issuance of earthquake predictions and warnings, in Earthquake Prediction Opportunity to Avert Disaster, U.S. GeoL Surv. Cir. 729, 10-12. Japanese National Land Policy Series (1978). Large-Scale Earthquakes Countermeasures Act, Law No. 73. - - - _. _ _ Seismological Society of America Bulletin (1983). Guidelines for earthquake predictors, 73, 1955-1956. South Coast Air Quality Management District, El Monte, California (1981). Air pollution episodes: what _ they are, What to do, 16 pp. Swacson, D. A,, T. J. Casadevall, D. Dzurisin, S. D. Malone, C. G. Newhall, and C. S. Weaver (1983). Predicting eruptions at Mount St. Helens, June 1980-December 1982, Science 221,1369-1376. U.S. Geological Survey (1977). Warning and preparedness for geologic-related hazards, Federal Register 42,19292-19296. U.S. National Research Council, Panel on the Public Policy Implications of Earthquake Prediction of the Advisory Committee on Emergency Planning (1975). Earthquake Prediction and Public Policy, U.S. National Academy of Sciences, Washington, D.CM 142 pp. U.S. National Research Council, Panel on Earthquake Prediction of the Committee on Seismology (1976). A Scientific and Technical Evaluation with Implications for Society, U.S. National Acad emy of Sciences, Washington, D.C., 62 pp. Wallace, R. E, (1981). Active faults, paleoseismology, and earthquake hazards in the western United States, in Earthquake Prediction, D. W. Simpson and P. G. Richards, Editors, Maurice Ewing Series 4, American Geophysical Union. U.S. GEOLOGICAL SURVEY C. F. RICHTER SEISMOLOGICAL LABORATORY 345 MlDDLEFIELD ROAD UNIVERSITY OF CALIFORNIA MEXLO PARK, CALIFORNIA 94025 (R.E.W.) SANTA CRUZ, CALIFORNIA 95064 (K.C.M.) CALIFORNIA DIVISION OF MINES AND GEOLOGY SACRAMENTO, CALIFORNIA 95814 (JJ.D.) Manuscript received 21 October 1983 79 APPENDIX 0 USGS, Terminology for Geologic Hazards Warnings 80 Geological Survey Terminology for Geologic Hazard Warnings SUMMARY: This notice describes changes in the terms and criteria used by the U.S. Geological Survey for issuing statements concerning geologic- related hazards to public officials and the public. For the purpose of this statement, a geologic hazard is a geologic condition, process, or potential event, such as an earthquake, volcanic eruption, or landslide, that poses a threat to the health, safety, or welfare of the public or to the functions or economy of a community or larger governmental entity. In this context a Geologic Hazard Warning is a formal statement by the Director of the U.S. Geological Survey that discusses a specific geologic condition, process, or potential event that poses a significant threat to the public, and for which some timely response would be expected. Directives or advisories to the public to take action, based on a Geologic Hazard Warning, may be issued by officials of State and local governments, and other Federal agencies, with authority and responsibility to use such statements. The term Hazard Warning is reserved for those situations posing a risk greater than normal and warranting considerations of a timely response in order to provide for public safety. Information regarding hazardous conditions that do not meet the criteria for a Hazard Warning may, however, also be sent to public officials as it becomes available. Transmittal of such information would not constitute a Hazard Warning. 1. The criteria for a Geologic Hazard Warning are: a. a degree of risk greater than normal for the area; or a hazardous condition that has recently developed or has only been recently recognized; and b. a threat that warrants consideration of a near-term public response. 2. A Geologic Hazard Warning consists of: a. a description of the geologic or other pertinent conditions that cause the concern; b. factors that indicate that such conditions constitute a potential hazard; c. location or area that may be affected; d. estimated severity and time of occurrence, if such estimates are justified by available information; 81 e. if possible, a probabilistic statement on the likelihood of a given event or events within a specified time period; and f. a description of continued Geological Survey involvement and estimate of what and when additional information might be available. If a life or property-threaten ing event is thought to be imminent, and immediate response is warranted by the public and public officials, the emergency nature of the Hazard Warning will be stated clearly either in the heading or the first sentence of the text of the warning statement. If the immediate crisis passes, either with or without the anticipated event, a revised statement will be issued to reflect the changed conditions and a re-evaluation of the geologic hazard. These changes in the terms and criteria do not entail or imply any changes to the procedures the U.S. Geological Survey uses to notify State and local governments, other Federal agencies, the public, or the news agenci.es and services. SUPPLEMENTARY INFORMATION: The Federal Register of April 12, 1977, Vol. 42, No. 70, pages 19292 to 19296 describes the previous terminology as well as the U.S. Geological Survey's authority to issue warnings of geologic- related hazards, capabilities to predict hazardous events, and provisional procedures to report hazardous conditions. Revised from Federal Register of January 31, 1984, Vol. 49, No. 21, pages 3938-3939.